agora inbox for [email protected]help / color / mirror / Atom feed
[PATCH v4] Make End-Of-Recovery error less scary 56+ messages / 3 participants [nested] [flat]
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v2] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 72 ++++++++++++++++++++------- src/backend/replication/walreceiver.c | 3 +- 2 files changed, 55 insertions(+), 20 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index d19408b3be..849cf6fe6b 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4282,12 +4282,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4295,14 +4298,16 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we are fetching checkpoint, we emit the error message right + * now. Otherwise the error is regarded as "end of WAL" and the + * message if any is shown as a part of the end-of-WAL message + * below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (fetching_ckpt && errormsg) + { + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); + } } /* @@ -4332,11 +4337,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else - { - /* No valid record available from this source */ - lastSourceFailed = true; + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) + { /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4349,11 +4355,18 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + (uint32) (ErrRecPtr >> 32), + (uint32) ErrRecPtr, + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4391,12 +4404,35 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg (fmt, (uint32) (EndRecPtr >> 32), + (uint32) EndRecPtr, + ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index 2ab15c3cbb..682dbb4e1f 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -478,8 +478,7 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, (uint32) (LogstreamResult.Write >> 32), (uint32) LogstreamResult.Write))); endofwal = true; -- 2.18.2 ----Next_Part(Thu_Mar__5_16_06_50_2020_335)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v17] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlogreader.c | 145 +++++++++++++++++----- src/backend/access/transam/xlogrecovery.c | 92 ++++++++++---- src/backend/replication/walreceiver.c | 7 +- src/bin/pg_waldump/pg_waldump.c | 13 +- src/include/access/xlogreader.h | 1 + src/test/recovery/t/011_crash_recovery.pl | 106 ++++++++++++++++ 6 files changed, 306 insertions(+), 58 deletions(-) diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index e437c42992..0942265408 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -46,6 +46,8 @@ static int ReadPageInternal(XLogReaderState *state, XLogRecPtr pageptr, int reqLen); static void XLogReaderInvalReadState(XLogReaderState *state); static XLogPageReadResult XLogDecodeNextRecord(XLogReaderState *state, bool non_blocking); +static bool ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record); static bool ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess); static bool ValidXLogRecord(XLogReaderState *state, XLogRecord *record, @@ -147,6 +149,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -552,6 +555,7 @@ XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking) /* reset error state */ state->errormsg_buf[0] = '\0'; decoded = NULL; + state->EndOfWAL = false; state->abortedRecPtr = InvalidXLogRecPtr; state->missingContrecPtr = InvalidXLogRecPtr; @@ -633,25 +637,21 @@ restart: Assert(pageHeaderSize <= readOff); /* - * Read the record length. + * Validate the record header. * - * NB: Even though we use an XLogRecord pointer here, the whole record - * header might not fit on this page. xl_tot_len is the first field of the - * struct, so it must be on this page (the records are MAXALIGNed), but we - * cannot access any other fields until we've verified that we got the - * whole header. - */ - record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); - total_len = record->xl_tot_len; - - /* - * If the whole record header is on this page, validate it immediately. - * Otherwise do just a basic sanity check on xl_tot_len, and validate the - * rest of the header after reading it from the next page. The xl_tot_len + * Even though we use an XLogRecord pointer here, the whole record header + * might not fit on this page. If the whole record header is on this page, + * validate it immediately. Even otherwise xl_tot_len must be on this page + * (it is the first field of MAXALIGNed records), but we still cannot + * access any further fields until we've verified that we got the whole + * header, so do just a basic sanity check on record length, and validate + * the rest of the header after reading it from the next page. The length * check is necessary here to ensure that we enter the "Need to reassemble * record" code path below; otherwise we might fail to apply * ValidXLogRecordHeader at all. */ + record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); + if (targetRecOff <= XLOG_BLCKSZ - SizeOfXLogRecord) { if (!ValidXLogRecordHeader(state, RecPtr, state->DecodeRecPtr, record, @@ -661,18 +661,14 @@ restart: } else { - /* XXX: more validation should be done here */ - if (total_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: wanted %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, total_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) goto err; - } + gotheader = false; } + total_len = record->xl_tot_len; + /* * Find space to decode this record. Don't allow oversized allocation if * the caller requested nonblocking. Otherwise, we *have* to try to @@ -904,6 +900,15 @@ err: */ state->abortedRecPtr = RecPtr; state->missingContrecPtr = targetPagePtr; + + /* + * If the message is not set yet, that means we failed to load the + * page for the record. Otherwise do not hide the existing message. + */ + if (state->errormsg_buf[0] == '\0') + report_invalid_record(state, + "missing contrecord at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); } if (decoded && decoded->oversized) @@ -1083,6 +1088,60 @@ XLogReaderInvalReadState(XLogReaderState *state) state->readLen = 0; } +/* + * Validate record length of an XLOG record header. + * + * This is substantially a part of ValidXLogRecordHeader. But XLogReadRecord + * needs this separate from the function in case of a partial record header. + */ +static bool +ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record) +{ + if (record->xl_tot_len == 0) + { + char *p; + char *pe; + + /* + * We are almost sure reaching the end of WAL, make sure that the + * whole page after the record is filled with zeroes. + */ + p = (char *) record; + pe = p + XLOG_BLCKSZ - (RecPtr & (XLOG_BLCKSZ - 1)); + + while (*p == 0 && p < pe) + p++; + + if (p == pe) + { + /* + * The page after the record is completely zeroed. That suggests + * we don't have a record after this point. We don't bother + * checking the pages after since they are not zeroed in the case + * of recycled segments. + */ + report_invalid_record(state, "empty record at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + return false; + } + } + + if (record->xl_tot_len < SizeOfXLogRecord) + { + report_invalid_record(state, + "invalid record length at %X/%X: wanted %u, got %u", + LSN_FORMAT_ARGS(RecPtr), + (uint32) SizeOfXLogRecord, record->xl_tot_len); + return false; + } + + return true; +} + /* * Validate an XLOG record header. * @@ -1094,14 +1153,9 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { - if (record->xl_tot_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: wanted %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, record->xl_tot_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) return false; - } + if (record->xl_rmid > RM_MAX_ID) { report_invalid_record(state, @@ -1200,6 +1254,31 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, XLogSegNoOffsetToRecPtr(segno, offset, state->segcxt.ws_segsize, recaddr); + StaticAssertStmt(XLOG_PAGE_MAGIC != 0, "XLOG_PAGE_MAGIC is zero"); + + if (hdr->xlp_magic == 0) + { + /* Regard an empty page as End-Of-WAL */ + int i; + + for (i = 0; i < XLOG_BLCKSZ && phdr[i] == 0; i++); + if (i == XLOG_BLCKSZ) + { + char fname[MAXFNAMELEN]; + + XLogFileName(fname, state->seg.ws_tli, segno, + state->segcxt.ws_segsize); + + report_invalid_record(state, + "empty page in log segment %s, offset %u", + fname, + offset); + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid magic number */ + } if (hdr->xlp_magic != XLOG_PAGE_MAGIC) { char fname[MAXFNAMELEN]; @@ -1285,6 +1364,14 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, LSN_FORMAT_ARGS(hdr->xlp_pageaddr), fname, offset); + + /* + * If the page address is less than expected we assume it is an unused + * page in a recycled segment. + */ + if (hdr->xlp_pageaddr < recaddr) + state->EndOfWAL = true; + return false; } diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index 9feea3e6ec..98382d66a4 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -1592,7 +1592,7 @@ PerformWalRecovery(void) /* just have to read next record after CheckPoint */ Assert(xlogreader->ReadRecPtr == CheckPointLoc); replayTLI = CheckPointTLI; - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } if (record != NULL) @@ -1706,7 +1706,7 @@ PerformWalRecovery(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } while (record != NULL); /* @@ -1765,13 +1765,20 @@ PerformWalRecovery(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that indicate @@ -2939,6 +2946,7 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); if (record == NULL) @@ -2954,6 +2962,18 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * EndRecPtr is the LSN we tried to read but failed. In the + * case of decoding error, it is at the end of the failed + * record but we don't have a means for now to know EndRecPtr + * is pointing to which of the beginning or ending of the + * failed record. + */ + ErrRecPtr = xlogreader->EndRecPtr; } if (readFile >= 0) @@ -2963,13 +2983,16 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * We only end up here without a message when XLogPageRead() failed + * in that case we already logged something, or just met end-of-WAL + * conditions. In StandbyMode that only happens if we have been + * triggered, so we shouldn't loop anymore in that case. When + * EndOfWAL is true, we don't emit that error if any immediately + * and instead will show it as a part of a decent end-of-wal + * message later. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -3000,11 +3023,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + Assert(ErrRecPtr != InvalidXLogRecPtr); + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -3017,11 +3043,16 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in pg_wal during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -3042,12 +3073,24 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * recovery ended. + * + * Emit a decent message if we met end-of-WAL. Otherwise we should + * have already emitted an error message. + */ + if (xlogreader->EndOfWAL) + ereport(LOG, + (errmsg("reached end of WAL at %X/%X on timeline %u", + LSN_FORMAT_ARGS(ErrRecPtr), replayTLI), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -3129,12 +3172,16 @@ retry: private->replayTLI, xlogreader->EndRecPtr)) { + Assert(!StandbyMode || CheckForStandbyTrigger()); + if (readFile >= 0) close(readFile); readFile = -1; readLen = 0; readSource = XLOG_FROM_ANY; + /* promotion exit is not end-of-WAL */ + xlogreader->EndOfWAL = !StandbyMode; return -1; } } @@ -3767,7 +3814,8 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + /* use currentSource as readSource is reset at failure */ + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index ceaff097b9..4f117ea4da 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -472,10 +472,9 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", - startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); + (errmsg("replication terminated by primary server at %X/%X on timeline %u.", + LSN_FORMAT_ARGS(LogstreamResult.Write), + startpointTLI))); endofwal = true; break; } diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index fc081adfb8..9bebca8154 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -1174,9 +1174,16 @@ main(int argc, char **argv) exit(0); if (errormsg) - fatal_error("error in WAL record at %X/%X: %s", - LSN_FORMAT_ARGS(xlogreader_state->ReadRecPtr), - errormsg); + { + if (xlogreader_state->EndOfWAL) + pg_log_info("end of WAL at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + else + fatal_error("error in WAL record at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + } XLogReaderFree(xlogreader_state); diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index f4388cc9be..21a8f9552c 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -201,6 +201,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* was the last attempt EOW? */ /* * Set at the end of recovery: the start point of a partial record at the diff --git a/src/test/recovery/t/011_crash_recovery.pl b/src/test/recovery/t/011_crash_recovery.pl index 14154d1ce0..01033334d6 100644 --- a/src/test/recovery/t/011_crash_recovery.pl +++ b/src/test/recovery/t/011_crash_recovery.pl @@ -10,7 +10,9 @@ use PostgreSQL::Test::Cluster; use PostgreSQL::Test::Utils; use Test::More; use Config; +use IPC::Run; +my $reached_eow_pat = "reached end of WAL at "; my $node = PostgreSQL::Test::Cluster->new('primary'); $node->init(allows_streaming => 1); $node->start; @@ -48,7 +50,15 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), # Crash and restart the postmaster $node->stop('immediate'); +my $logstart = get_log_size($node); $node->start; +my $max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, $reached_eow_pat, $logstart)); + sleep 0.5; +} +ok ($max_attempts >= 0, "end-of-wal is logged"); # Make sure we really got a new xid cmp_ok($node->safe_psql('postgres', 'SELECT pg_current_xact_id()'), @@ -61,4 +71,100 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), $stdin .= "\\q\n"; $tx->finish; # wait for psql to quit gracefully +my $segsize = $node->safe_psql('postgres', + qq[SELECT setting FROM pg_settings WHERE name = 'wal_segment_size';]); + +# make sure no records afterwards go to the next segment +$node->safe_psql('postgres', qq[ + SELECT pg_switch_wal(); + CHECKPOINT; + CREATE TABLE t(); +]); +$node->stop('immediate'); + +# identify REDO WAL file +my $cmd = "pg_controldata -D " . $node->data_dir(); +$cmd = ['pg_controldata', '-D', $node->data_dir()]; +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stdout =~ /^Latest checkpoint's REDO WAL file:[ \t] *(.+)$/m, + "checkpoint file is identified"); +my $chkptfile = $1; + +# identify the last record +my $walfile = $node->data_dir() . "/pg_wal/$chkptfile"; +$cmd = ['pg_waldump', $walfile]; +$stdout = ''; +$stderr = ''; +my $lastrec; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +foreach my $l (split(/\r?\n/, $stdout)) +{ + $lastrec = $l; +} +ok(defined $lastrec, "last WAL record is extracted"); +ok($stderr =~ /end of WAL at ([0-9A-F\/]+): .* at \g1/, + "pg_waldump emits the correct ending message"); + +# read the last record LSN excluding leading zeroes +ok ($lastrec =~ /, lsn: 0\/0*([1-9A-F][0-9A-F]+),/, + "LSN of the last record identified"); +my $lastlsn = $1; + +# corrupt the last record +my $offset = hex($lastlsn) % $segsize; +open(my $segf, '+<', $walfile) or die "failed to open $walfile\n"; +seek($segf, $offset, 0); # halfway break the last record +print $segf "\0\0\0\0"; +close($segf); + +# pg_waldump complains about the corrupted record +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stderr =~ /fatal: error in WAL record at 0\/$lastlsn: .* at 0\/$lastlsn/, + "pg_waldump emits the correct error message"); + +# also server complains +$logstart = get_log_size($node); +$node->start; +$max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, "WARNING: invalid record length at 0/$lastlsn: wanted [0-9]+, got 0", + $logstart)); + sleep 0.5; +} +ok($max_attempts >= 0, "header error is logged at $lastlsn"); + +# no end-of-wal message should be seen this time +ok(!find_in_log($node, $reached_eow_pat, $logstart), + "false log message is not emitted"); + +$node->stop('immediate'); + done_testing(); + +#### helper routines +# return the size of logfile of $node in bytes +sub get_log_size +{ + my ($node) = @_; + + return (stat $node->logfile)[7]; +} + +# find $pat in logfile of $node after $off-th byte +sub find_in_log +{ + my ($node, $pat, $off) = @_; + + $off = 0 unless defined $off; + my $log = PostgreSQL::Test::Utils::slurp_file($node->logfile); + return 0 if (length($log) <= $off); + + $log = substr($log, $off); + + return $log =~ m/$pat/; +} -- 2.27.0 ----Next_Part(Tue_Mar_22_11_34_46_2022_933)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v10] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlog.c | 91 +++++++++++++----- src/backend/access/transam/xlogreader.c | 77 +++++++++++++++ src/backend/replication/walreceiver.c | 7 +- src/bin/pg_waldump/pg_waldump.c | 13 ++- src/include/access/xlogreader.h | 1 + src/test/recovery/t/011_crash_recovery.pl | 110 +++++++++++++++++++++- 6 files changed, 269 insertions(+), 30 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index dfe2a0bcce..378c13ccf7 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4480,6 +4480,7 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); if (record == NULL) @@ -4495,6 +4496,18 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * NULL ReadRecPtr means we could not read a record at the + * beginning. In that case EndRecPtr is storing the LSN of the + * record we tried to read. + */ + ErrRecPtr = + xlogreader->ReadRecPtr ? + xlogreader->ReadRecPtr : xlogreader->EndRecPtr; } if (readFile >= 0) @@ -4504,13 +4517,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we get here for other than end-of-wal, emit the error + * message right now. Otherwise the message if any is shown as a + * part of the end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4541,11 +4553,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4558,11 +4571,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4610,12 +4629,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * If we haven't emit an error message, we have safely reached the + * end-of-WAL. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(ErrRecPtr), replayTLI, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7544,7 +7584,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } if (record != NULL) @@ -7782,7 +7822,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } while (record != NULL); /* @@ -7842,13 +7882,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -13097,7 +13144,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 35029cf97d..9bcc4a2d37 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -121,6 +121,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -292,6 +293,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); state->abortedRecPtr = InvalidXLogRecPtr; @@ -588,6 +590,15 @@ err: */ state->abortedRecPtr = RecPtr; state->missingContrecPtr = targetPagePtr; + + /* + * If the message is not set yet, that means we failed to load the + * page for the record. Otherwise do not hide the existing message. + */ + if (state->errormsg_buf[0] == '\0') + report_invalid_record(state, + "missing contrecord at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); } /* @@ -730,6 +741,39 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* + * We are almost sure reaching the end of WAL, make sure that the + * whole page after the record is filled with zeroes. + */ + char *p = (char *) record; + char *pe; + + /* set pe to the beginning of the next page */ + pe = (char *) record + XLOG_BLCKSZ - (RecPtr & (XLOG_BLCKSZ - 1)); + + while (*p == 0 && p < pe) + p++; + + if (p == pe) + { + /* + * The page after the record is completely zeroed. That suggests + * we don't have a record after this point. We don't bother + * checking the pages after since they are not zeroed in the case + * of recycled segments. + */ + report_invalid_record(state, "empty record found at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid record length */ + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, @@ -836,6 +880,31 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, XLogSegNoOffsetToRecPtr(segno, offset, state->segcxt.ws_segsize, recaddr); + StaticAssertStmt(XLOG_PAGE_MAGIC != 0, "XLOG_PAGE_MAGIC is zero"); + + if (hdr->xlp_magic == 0) + { + /* Regard an empty page as End-Of-WAL */ + int i; + + for (i = 0; i < XLOG_BLCKSZ && phdr[i] == 0; i++); + if (i == XLOG_BLCKSZ) + { + char fname[MAXFNAMELEN]; + + XLogFileName(fname, state->seg.ws_tli, segno, + state->segcxt.ws_segsize); + + report_invalid_record(state, + "empty page in log segment %s, offset %u", + fname, + offset); + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid magic number */ + } if (hdr->xlp_magic != XLOG_PAGE_MAGIC) { char fname[MAXFNAMELEN]; @@ -921,6 +990,14 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, LSN_FORMAT_ARGS(hdr->xlp_pageaddr), fname, offset); + + /* + * If the page address is less than expected we assume it is an unused + * page in a recycled segment. + */ + if (hdr->xlp_pageaddr < recaddr) + state->EndOfWAL = true; + return false; } diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index b39fce8c23..8e1fa32489 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -471,10 +471,9 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", - startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", + startpointTLI, + LSN_FORMAT_ARGS(LogstreamResult.Write)))); endofwal = true; break; } diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index a6251e1a96..3745e76488 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -1176,9 +1176,16 @@ main(int argc, char **argv) exit(0); if (errormsg) - fatal_error("error in WAL record at %X/%X: %s", - LSN_FORMAT_ARGS(xlogreader_state->ReadRecPtr), - errormsg); + { + if (xlogreader_state->EndOfWAL) + pg_log_info("end of WAL at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + else + fatal_error("error in WAL record at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + } XLogReaderFree(xlogreader_state); diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 477f0efe26..3eeba220a1 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* * Set at the end of recovery: the start point of a partial record at the diff --git a/src/test/recovery/t/011_crash_recovery.pl b/src/test/recovery/t/011_crash_recovery.pl index 3892aba3e5..67d264df26 100644 --- a/src/test/recovery/t/011_crash_recovery.pl +++ b/src/test/recovery/t/011_crash_recovery.pl @@ -10,9 +10,11 @@ use PostgreSQL::Test::Cluster; use PostgreSQL::Test::Utils; use Test::More; use Config; +use IPC::Run; -plan tests => 3; +plan tests => 11; +my $reached_eow_pat = "reached end of WAL at "; my $node = PostgreSQL::Test::Cluster->new('primary'); $node->init(allows_streaming => 1); $node->start; @@ -50,7 +52,15 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), # Crash and restart the postmaster $node->stop('immediate'); +my $logstart = get_log_size($node); $node->start; +my $max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, $reached_eow_pat, $logstart)); + sleep 0.5; +} +ok ($max_attempts >= 0, "end-of-wal is logged"); # Make sure we really got a new xid cmp_ok($node->safe_psql('postgres', 'SELECT pg_current_xact_id()'), @@ -62,3 +72,101 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), $stdin .= "\\q\n"; $tx->finish; # wait for psql to quit gracefully + +my $segsize = $node->safe_psql('postgres', + qq[SELECT setting FROM pg_settings WHERE name = 'wal_segment_size';]); + +# make sure no records afterwards go to the next segment +$node->safe_psql('postgres', qq[ + SELECT pg_switch_wal(); + CHECKPOINT; + CREATE TABLE t(); +]); +$node->stop('immediate'); + +# identify REDO WAL file +my $cmd = "pg_controldata -D " . $node->data_dir(); +my $chkptfile; +$cmd = ['pg_controldata', '-D', $node->data_dir()]; +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stdout =~ /^Latest checkpoint's REDO WAL file:[ \t] *(.+)$/m, + "checkpoint file is identified"); +my $chkptfile = $1; + +# identify the last record +my $walfile = $node->data_dir() . "/pg_wal/$chkptfile"; +$cmd = ['pg_waldump', $walfile]; +$stdout = ''; +$stderr = ''; +my $lastrec; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +foreach my $l (split(/\r?\n/, $stdout)) +{ + $lastrec = $l; +} +ok(defined $lastrec, "last WAL record is extracted"); +ok($stderr =~ /end of WAL at ([0-9A-F\/]+): .* at \g1/, + "pg_waldump emits the correct ending message"); + +# read the last record LSN excluding leading zeroes +ok ($lastrec =~ /, lsn: 0\/0*([1-9A-F][0-9A-F]+),/, + "LSN of the last record identified"); +my $lastlsn = $1; + +# corrupt the last record +my $offset = hex($lastlsn) % $segsize; +open(my $segf, '+<', $walfile) or die "failed to open $walfile\n"; +seek($segf, $offset, 0); # halfway break the last record +print $segf "\0\0\0\0"; +close($segf); + +# pg_waldump complains about the corrupted record +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stderr =~ /fatal: error in WAL record at 0\/$lastlsn: .* at 0\/$lastlsn/, + "pg_waldump emits the correct error message"); + +# also server complains +$logstart = get_log_size($node); +$node->start; +$max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, "WARNING: invalid record length at 0/$lastlsn: wanted [0-9]+, got 0", + $logstart)); + sleep 0.5; +} +ok($max_attempts >= 0, "header error is logged at $lastlsn"); + +# and the end-of-wal messages shouldn't be seen +# the same message has been confirmed in the past +ok(!find_in_log($node, $reached_eow_pat, $logstart), + "false log message is not emitted"); + +$node->stop('immediate'); + +#### helper routines +# return the size of logfile of $node in bytes +sub get_log_size +{ + my ($node) = @_; + + return (stat $node->logfile)[7]; +} + +# find $pat in logfile of $node after $off-th byte +sub find_in_log +{ + my ($node, $pat, $off) = @_; + + $off = 0 unless defined $off; + my $log = PostgreSQL::Test::Utils::slurp_file($node->logfile); + return 0 if (length($log) <= $off); + + $log = substr($log, $off); + + return $log =~ m/$pat/; +} -- 2.27.0 ----Next_Part(Tue_Feb__1_11_58_01_2022_779)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v8] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlog.c | 91 +++++++++++++----- src/backend/access/transam/xlogreader.c | 64 +++++++++++++ src/backend/replication/walreceiver.c | 3 +- src/bin/pg_waldump/pg_waldump.c | 13 ++- src/include/access/xlogreader.h | 1 + src/test/recovery/t/011_crash_recovery.pl | 110 +++++++++++++++++++++- 6 files changed, 254 insertions(+), 28 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 58922f7ede..c08b9554b3 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4480,6 +4480,7 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); if (record == NULL) @@ -4495,6 +4496,18 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * NULL ReadRecPtr means we could not read a record at the + * beginning. In that case EndRecPtr is storing the LSN of the + * record we tried to read. + */ + ErrRecPtr = + xlogreader->ReadRecPtr ? + xlogreader->ReadRecPtr : xlogreader->EndRecPtr; } if (readFile >= 0) @@ -4504,13 +4517,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we get here for other than end-of-wal, emit the error message + * right now. Otherwise the message if any is shown as a part of + * the end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4541,11 +4553,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4558,11 +4571,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4610,12 +4629,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * If we haven't emit an error message, we have safely reached the + * end-of-WAL. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(ErrRecPtr), replayTLI, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7544,7 +7584,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } if (record != NULL) @@ -7782,7 +7822,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } while (record != NULL); /* @@ -7842,13 +7882,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -13097,7 +13144,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 35029cf97d..55f54cd98d 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -121,6 +121,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -292,6 +293,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); state->abortedRecPtr = InvalidXLogRecPtr; @@ -588,6 +590,15 @@ err: */ state->abortedRecPtr = RecPtr; state->missingContrecPtr = targetPagePtr; + + /* + * If the message is not set yet, that means we failed to load the + * page for the record. Otherwise do not hide the existing message. + */ + if (state->errormsg_buf[0] == '\0') + report_invalid_record(state, + "missing contrecord at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); } /* @@ -730,6 +741,36 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* + * We are almost sure reaching the end of WAL, make sure that the whole + * header is zeroed. + */ + char *p = (char *)record; + char *pe = (char *)record + SizeOfXLogRecord; + + while (*p == 0 && p < pe) + p++; + + if (p == pe) + { + /* it is completely zeroed, call it a day */ + report_invalid_record(state, "empty record header found at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + } + else + { + /* Otherwise the header is corrupted. */ + report_invalid_record(state, "garbage record header at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + } + + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, @@ -836,6 +877,29 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, XLogSegNoOffsetToRecPtr(segno, offset, state->segcxt.ws_segsize, recaddr); + StaticAssertStmt(XLOG_PAGE_MAGIC != 0, "XLOG_PAGE_MAGIC is zero"); + + if (hdr->xlp_magic == 0) + { + /* Regard an empty page as End-Of-WAL */ + int i; + + for (i = 0 ; i < XLOG_BLCKSZ && phdr[i] == 0 ; i++); + if (i == XLOG_BLCKSZ) + { + char fname[MAXFNAMELEN]; + + XLogFileName(fname, state->seg.ws_tli, segno, + state->segcxt.ws_segsize); + + report_invalid_record(state, + "empty page in log segment %s, offset %u", + fname, + offset); + state->EndOfWAL = true; + return false; + } + } if (hdr->xlp_magic != XLOG_PAGE_MAGIC) { char fname[MAXFNAMELEN]; diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index b39fce8c23..3034f8281e 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -471,8 +471,7 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, LSN_FORMAT_ARGS(LogstreamResult.Write)))); endofwal = true; diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index a6251e1a96..3745e76488 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -1176,9 +1176,16 @@ main(int argc, char **argv) exit(0); if (errormsg) - fatal_error("error in WAL record at %X/%X: %s", - LSN_FORMAT_ARGS(xlogreader_state->ReadRecPtr), - errormsg); + { + if (xlogreader_state->EndOfWAL) + pg_log_info("end of WAL at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + else + fatal_error("error in WAL record at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + } XLogReaderFree(xlogreader_state); diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 477f0efe26..3eeba220a1 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* * Set at the end of recovery: the start point of a partial record at the diff --git a/src/test/recovery/t/011_crash_recovery.pl b/src/test/recovery/t/011_crash_recovery.pl index 3892aba3e5..b793280a5c 100644 --- a/src/test/recovery/t/011_crash_recovery.pl +++ b/src/test/recovery/t/011_crash_recovery.pl @@ -10,9 +10,11 @@ use PostgreSQL::Test::Cluster; use PostgreSQL::Test::Utils; use Test::More; use Config; +use IPC::Run; -plan tests => 3; +plan tests => 11; +my $reached_eow_pat = "reached end of WAL at "; my $node = PostgreSQL::Test::Cluster->new('primary'); $node->init(allows_streaming => 1); $node->start; @@ -50,7 +52,15 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), # Crash and restart the postmaster $node->stop('immediate'); +my $logstart = get_log_size($node); $node->start; +my $max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, $reached_eow_pat, $logstart)); + sleep 0.5; +} +ok ($max_attempts >= 0, "end-of-wal is logged"); # Make sure we really got a new xid cmp_ok($node->safe_psql('postgres', 'SELECT pg_current_xact_id()'), @@ -62,3 +72,101 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), $stdin .= "\\q\n"; $tx->finish; # wait for psql to quit gracefully + +my $segsize = $node->safe_psql('postgres', + qq[SELECT setting FROM pg_settings WHERE name = 'wal_segment_size';]); + +# make sure no records afterwards go to the next segment +$node->safe_psql('postgres', qq[ + SELECT pg_switch_wal(); + CHECKPOINT; + CREATE TABLE t(); +]); +$node->stop('immediate'); + +# identify REDO WAL file +my $cmd = "pg_controldata -D " . $node->data_dir(); +my $chkptfile; +$cmd = ['pg_controldata', '-D', $node->data_dir()]; +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stdout =~ /^Latest checkpoint's REDO WAL file:[ \t] *(.+)$/m, + "checkpoint file is identified"); +my $chkptfile = $1; + +# identify the last record +my $walfile = $node->data_dir() . "/pg_wal/$chkptfile"; +$cmd = ['pg_waldump', $walfile]; +$stdout = ''; +$stderr = ''; +my $lastrec; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +foreach my $l (split(/\r?\n/, $stdout)) +{ + $lastrec = $l; +} +ok(defined $lastrec, "last WAL record is extracted"); +ok($stderr =~ /end of WAL at ([0-9A-F\/]+): .* at \g1/, + "pg_waldump emits the correct ending message"); + +# read the last record LSN excluding leading zeroes +ok ($lastrec =~ /, lsn: 0\/0*([1-9A-F][0-9A-F]+),/, + "LSN of the last record identified"); +my $lastlsn = $1; + +# corrupt the last record +my $offset = hex($lastlsn) % $segsize; +open(my $segf, '+<', $walfile) or die "failed to open $walfile\n"; +seek($segf, $offset, 0); # halfway break the last record +print $segf "\0\0\0\0"; +close($segf); + +# pg_waldump complains about the corrupted record +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stderr =~ /fatal: error in WAL record at 0\/$lastlsn: .* at 0\/$lastlsn/, + "pg_waldump emits the correct error message"); + +# also server complains +$logstart = get_log_size($node); +$node->start; +$max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, "WARNING: garbage record header at 0/$lastlsn", + $logstart)); + sleep 0.5; +} +ok($max_attempts >= 0, "header error is logged at $lastlsn"); + +# and the end-of-wal messages shouldn't be seen +# the same message has been confirmed in the past +ok(!find_in_log($node, $reached_eow_pat, $logstart), + "false log message is not emitted"); + +$node->stop('immediate'); + +#### helper routines +# return the size of logfile of $node in bytes +sub get_log_size +{ + my ($node) = @_; + + return (stat $node->logfile)[7]; +} + +# find $pat in logfile of $node after $off-th byte +sub find_in_log +{ + my ($node, $pat, $off) = @_; + + $off = 0 unless defined $off; + my $log = PostgreSQL::Test::Utils::slurp_file($node->logfile); + return 0 if (length($log) <= $off); + + $log = substr($log, $off); + + return $log =~ m/$pat/; +} -- 2.27.0 ----Next_Part(Tue_Jan_25_17_34_56_2022_438)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v6] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlog.c | 89 +++++++++++++++++++------ src/backend/access/transam/xlogreader.c | 42 ++++++++++++ src/backend/replication/walreceiver.c | 3 +- src/include/access/xlogreader.h | 1 + 4 files changed, 111 insertions(+), 24 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 5cda30836f..e90c69810b 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4477,6 +4477,7 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; @@ -4494,6 +4495,16 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * NULL ReadRecPtr means we could not read a record at + * beginning. In that case EndRecPtr is storing the LSN of the + * record we tried to read. + */ + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; } if (readFile >= 0) @@ -4503,13 +4514,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we get here for other than end-of-wal, emit the error message + * right now. Otherwise the message if any is shown as a part of + * the end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4540,11 +4550,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4557,11 +4568,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4609,12 +4626,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * If we haven't emit an error message, we have safely reached the + * end-of-WAL. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(ErrRecPtr), replayTLI, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7544,7 +7582,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false, ThisTimeLineID); + record = ReadRecord(xlogreader, WARNING, false, ThisTimeLineID); } if (record != NULL) @@ -7781,7 +7819,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false, ThisTimeLineID); + record = ReadRecord(xlogreader, WARNING, false, ThisTimeLineID); } while (record != NULL); /* @@ -7841,13 +7879,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -13147,7 +13192,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index f39f8044a9..273b927cd9 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -121,6 +121,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -292,6 +293,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); state->abortedRecPtr = InvalidXLogRecPtr; @@ -588,6 +590,16 @@ err: */ state->abortedRecPtr = RecPtr; state->missingContrecPtr = targetPagePtr; + + /* + * If the messages is not set yet, that means we failed to load the + * page for the record. Otherwise do not hide the existing message at + * it should be more detailed. + */ + if (state->errormsg_buf[0] == '\0') + report_invalid_record(state, + "missing contrecord at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); } /* @@ -730,6 +742,36 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* + * We are almost sure reaching the end of WAL, make sure that the whole + * header is zeroed. + */ + char *p = (char *)record; + char *pe = (char *)record + SizeOfXLogRecord; + + while (*p == 0 && p < pe) + p++; + + if (p == pe) + { + /* it is completely zeroed, call it a day */ + report_invalid_record(state, "empty record header found at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + } + else + { + /* Otherwise we found a garbage header.. */ + report_invalid_record(state, "garbage record header at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + } + + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index 7a7eb3784e..ba3c4bd550 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -471,8 +471,7 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, LSN_FORMAT_ARGS(LogstreamResult.Write)))); endofwal = true; diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index de6fd791fe..1241b85838 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* * Set at the end of recovery: the start point of a partial record at the -- 2.27.0 ----Next_Part(Tue_Nov__9_16_27_52_2021_065)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v9] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlog.c | 91 +++++++++++++----- src/backend/access/transam/xlogreader.c | 61 ++++++++++++ src/backend/replication/walreceiver.c | 3 +- src/bin/pg_waldump/pg_waldump.c | 13 ++- src/include/access/xlogreader.h | 1 + src/test/recovery/t/011_crash_recovery.pl | 110 +++++++++++++++++++++- 6 files changed, 251 insertions(+), 28 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index dfe2a0bcce..5727e0939f 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4480,6 +4480,7 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); if (record == NULL) @@ -4495,6 +4496,18 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * NULL ReadRecPtr means we could not read a record at the + * beginning. In that case EndRecPtr is storing the LSN of the + * record we tried to read. + */ + ErrRecPtr = + xlogreader->ReadRecPtr ? + xlogreader->ReadRecPtr : xlogreader->EndRecPtr; } if (readFile >= 0) @@ -4504,13 +4517,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we get here for other than end-of-wal, emit the error message + * right now. Otherwise the message if any is shown as a part of + * the end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4541,11 +4553,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4558,11 +4571,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4610,12 +4629,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * If we haven't emit an error message, we have safely reached the + * end-of-WAL. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(ErrRecPtr), replayTLI, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7544,7 +7584,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } if (record != NULL) @@ -7782,7 +7822,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } while (record != NULL); /* @@ -7842,13 +7882,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -13097,7 +13144,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 35029cf97d..418fb66ef2 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -121,6 +121,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -292,6 +293,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); state->abortedRecPtr = InvalidXLogRecPtr; @@ -588,6 +590,15 @@ err: */ state->abortedRecPtr = RecPtr; state->missingContrecPtr = targetPagePtr; + + /* + * If the message is not set yet, that means we failed to load the + * page for the record. Otherwise do not hide the existing message. + */ + if (state->errormsg_buf[0] == '\0') + report_invalid_record(state, + "missing contrecord at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); } /* @@ -730,6 +741,31 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* + * We are almost sure reaching the end of WAL, make sure that the whole + * header is filled with zeroes. + */ + char *p = (char *)record; + char *pe = (char *)record + SizeOfXLogRecord; + + while (*p == 0 && p < pe) + p++; + + if (p == pe) + { + /* it is completely zeroed, call it a day */ + report_invalid_record(state, "empty record header found at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid record length */ + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, @@ -836,6 +872,31 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, XLogSegNoOffsetToRecPtr(segno, offset, state->segcxt.ws_segsize, recaddr); + StaticAssertStmt(XLOG_PAGE_MAGIC != 0, "XLOG_PAGE_MAGIC is zero"); + + if (hdr->xlp_magic == 0) + { + /* Regard an empty page as End-Of-WAL */ + int i; + + for (i = 0 ; i < XLOG_BLCKSZ && phdr[i] == 0 ; i++); + if (i == XLOG_BLCKSZ) + { + char fname[MAXFNAMELEN]; + + XLogFileName(fname, state->seg.ws_tli, segno, + state->segcxt.ws_segsize); + + report_invalid_record(state, + "empty page in log segment %s, offset %u", + fname, + offset); + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid magic number */ + } if (hdr->xlp_magic != XLOG_PAGE_MAGIC) { char fname[MAXFNAMELEN]; diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index b39fce8c23..3034f8281e 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -471,8 +471,7 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, LSN_FORMAT_ARGS(LogstreamResult.Write)))); endofwal = true; diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index a6251e1a96..3745e76488 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -1176,9 +1176,16 @@ main(int argc, char **argv) exit(0); if (errormsg) - fatal_error("error in WAL record at %X/%X: %s", - LSN_FORMAT_ARGS(xlogreader_state->ReadRecPtr), - errormsg); + { + if (xlogreader_state->EndOfWAL) + pg_log_info("end of WAL at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + else + fatal_error("error in WAL record at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + } XLogReaderFree(xlogreader_state); diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 477f0efe26..3eeba220a1 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* * Set at the end of recovery: the start point of a partial record at the diff --git a/src/test/recovery/t/011_crash_recovery.pl b/src/test/recovery/t/011_crash_recovery.pl index 3892aba3e5..67d264df26 100644 --- a/src/test/recovery/t/011_crash_recovery.pl +++ b/src/test/recovery/t/011_crash_recovery.pl @@ -10,9 +10,11 @@ use PostgreSQL::Test::Cluster; use PostgreSQL::Test::Utils; use Test::More; use Config; +use IPC::Run; -plan tests => 3; +plan tests => 11; +my $reached_eow_pat = "reached end of WAL at "; my $node = PostgreSQL::Test::Cluster->new('primary'); $node->init(allows_streaming => 1); $node->start; @@ -50,7 +52,15 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), # Crash and restart the postmaster $node->stop('immediate'); +my $logstart = get_log_size($node); $node->start; +my $max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, $reached_eow_pat, $logstart)); + sleep 0.5; +} +ok ($max_attempts >= 0, "end-of-wal is logged"); # Make sure we really got a new xid cmp_ok($node->safe_psql('postgres', 'SELECT pg_current_xact_id()'), @@ -62,3 +72,101 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), $stdin .= "\\q\n"; $tx->finish; # wait for psql to quit gracefully + +my $segsize = $node->safe_psql('postgres', + qq[SELECT setting FROM pg_settings WHERE name = 'wal_segment_size';]); + +# make sure no records afterwards go to the next segment +$node->safe_psql('postgres', qq[ + SELECT pg_switch_wal(); + CHECKPOINT; + CREATE TABLE t(); +]); +$node->stop('immediate'); + +# identify REDO WAL file +my $cmd = "pg_controldata -D " . $node->data_dir(); +my $chkptfile; +$cmd = ['pg_controldata', '-D', $node->data_dir()]; +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stdout =~ /^Latest checkpoint's REDO WAL file:[ \t] *(.+)$/m, + "checkpoint file is identified"); +my $chkptfile = $1; + +# identify the last record +my $walfile = $node->data_dir() . "/pg_wal/$chkptfile"; +$cmd = ['pg_waldump', $walfile]; +$stdout = ''; +$stderr = ''; +my $lastrec; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +foreach my $l (split(/\r?\n/, $stdout)) +{ + $lastrec = $l; +} +ok(defined $lastrec, "last WAL record is extracted"); +ok($stderr =~ /end of WAL at ([0-9A-F\/]+): .* at \g1/, + "pg_waldump emits the correct ending message"); + +# read the last record LSN excluding leading zeroes +ok ($lastrec =~ /, lsn: 0\/0*([1-9A-F][0-9A-F]+),/, + "LSN of the last record identified"); +my $lastlsn = $1; + +# corrupt the last record +my $offset = hex($lastlsn) % $segsize; +open(my $segf, '+<', $walfile) or die "failed to open $walfile\n"; +seek($segf, $offset, 0); # halfway break the last record +print $segf "\0\0\0\0"; +close($segf); + +# pg_waldump complains about the corrupted record +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stderr =~ /fatal: error in WAL record at 0\/$lastlsn: .* at 0\/$lastlsn/, + "pg_waldump emits the correct error message"); + +# also server complains +$logstart = get_log_size($node); +$node->start; +$max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, "WARNING: invalid record length at 0/$lastlsn: wanted [0-9]+, got 0", + $logstart)); + sleep 0.5; +} +ok($max_attempts >= 0, "header error is logged at $lastlsn"); + +# and the end-of-wal messages shouldn't be seen +# the same message has been confirmed in the past +ok(!find_in_log($node, $reached_eow_pat, $logstart), + "false log message is not emitted"); + +$node->stop('immediate'); + +#### helper routines +# return the size of logfile of $node in bytes +sub get_log_size +{ + my ($node) = @_; + + return (stat $node->logfile)[7]; +} + +# find $pat in logfile of $node after $off-th byte +sub find_in_log +{ + my ($node, $pat, $off) = @_; + + $off = 0 unless defined $off; + my $log = PostgreSQL::Test::Utils::slurp_file($node->logfile); + return 0 if (length($log) <= $off); + + $log = substr($log, $off); + + return $log =~ m/$pat/; +} -- 2.27.0 ----Next_Part(Thu_Jan_27_10_35_47_2022_592)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v7] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlog.c | 89 +++++++++++++++++++------ src/backend/access/transam/xlogreader.c | 42 ++++++++++++ src/backend/replication/walreceiver.c | 3 +- src/include/access/xlogreader.h | 1 + 4 files changed, 112 insertions(+), 23 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index d894af310a..fa435faec4 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4469,6 +4469,7 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); if (record == NULL) @@ -4484,6 +4485,18 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * NULL ReadRecPtr means we could not read a record at + * beginning. In that case EndRecPtr is storing the LSN of the + * record we tried to read. + */ + ErrRecPtr = + xlogreader->ReadRecPtr ? + xlogreader->ReadRecPtr : xlogreader->EndRecPtr; } if (readFile >= 0) @@ -4493,12 +4506,11 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we get here for other than end-of-wal, emit the error message + * right now. Otherwise the message if any is shown as a part of + * the end-of-WAL message below. */ - if (errormsg) + if (!xlogreader->EndOfWAL && errormsg) ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4530,11 +4542,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4547,11 +4560,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4599,12 +4618,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * If we haven't emit an error message, we have safely reached the + * end-of-WAL. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(ErrRecPtr), replayTLI, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7536,7 +7576,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } if (record != NULL) @@ -7774,7 +7814,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } while (record != NULL); /* @@ -7834,13 +7874,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -13130,7 +13177,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 3a7de02565..e16b6fe041 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -121,6 +121,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -292,6 +293,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); state->abortedRecPtr = InvalidXLogRecPtr; @@ -588,6 +590,16 @@ err: */ state->abortedRecPtr = RecPtr; state->missingContrecPtr = targetPagePtr; + + /* + * If the messages is not set yet, that means we failed to load the + * page for the record. Otherwise do not hide the existing message at + * it should be more detailed. + */ + if (state->errormsg_buf[0] == '\0') + report_invalid_record(state, + "missing contrecord at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); } /* @@ -730,6 +742,36 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* + * We are almost sure reaching the end of WAL, make sure that the whole + * header is zeroed. + */ + char *p = (char *)record; + char *pe = (char *)record + SizeOfXLogRecord; + + while (*p == 0 && p < pe) + p++; + + if (p == pe) + { + /* it is completely zeroed, call it a day */ + report_invalid_record(state, "empty record header found at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + } + else + { + /* Otherwise we found a garbage header.. */ + report_invalid_record(state, "garbage record header at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + } + + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index 7a7eb3784e..ba3c4bd550 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -471,8 +471,7 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, LSN_FORMAT_ARGS(LogstreamResult.Write)))); endofwal = true; diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index de6fd791fe..1241b85838 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* * Set at the end of recovery: the start point of a partial record at the -- 2.27.0 ----Next_Part(Wed_Dec__8_16_01_47_2021_480)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v14] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlogreader.c | 78 ++++++++++++++++ src/backend/access/transam/xlogrecovery.c | 92 ++++++++++++++----- src/backend/replication/walreceiver.c | 7 +- src/bin/pg_waldump/pg_waldump.c | 13 ++- src/include/access/xlogreader.h | 1 + src/test/recovery/t/011_crash_recovery.pl | 106 ++++++++++++++++++++++ 6 files changed, 268 insertions(+), 29 deletions(-) diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 35029cf97d..22982c4de7 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -121,6 +121,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -292,6 +293,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); state->abortedRecPtr = InvalidXLogRecPtr; @@ -588,6 +590,15 @@ err: */ state->abortedRecPtr = RecPtr; state->missingContrecPtr = targetPagePtr; + + /* + * If the message is not set yet, that means we failed to load the + * page for the record. Otherwise do not hide the existing message. + */ + if (state->errormsg_buf[0] == '\0') + report_invalid_record(state, + "missing contrecord at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); } /* @@ -730,6 +741,40 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* + * We are almost sure reaching the end of WAL, make sure that the + * whole page after the record is filled with zeroes. + */ + char *p; + char *pe; + + /* scan from the beginning of the record to the end of block */ + p = (char *) record; + pe = p + XLOG_BLCKSZ - (RecPtr & (XLOG_BLCKSZ - 1)); + + while (*p == 0 && p < pe) + p++; + + if (p == pe) + { + /* + * The page after the record is completely zeroed. That suggests + * we don't have a record after this point. We don't bother + * checking the pages after since they are not zeroed in the case + * of recycled segments. + */ + report_invalid_record(state, "empty record at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid record length */ + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, @@ -836,6 +881,31 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, XLogSegNoOffsetToRecPtr(segno, offset, state->segcxt.ws_segsize, recaddr); + StaticAssertStmt(XLOG_PAGE_MAGIC != 0, "XLOG_PAGE_MAGIC is zero"); + + if (hdr->xlp_magic == 0) + { + /* Regard an empty page as End-Of-WAL */ + int i; + + for (i = 0; i < XLOG_BLCKSZ && phdr[i] == 0; i++); + if (i == XLOG_BLCKSZ) + { + char fname[MAXFNAMELEN]; + + XLogFileName(fname, state->seg.ws_tli, segno, + state->segcxt.ws_segsize); + + report_invalid_record(state, + "empty page in log segment %s, offset %u", + fname, + offset); + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid magic number */ + } if (hdr->xlp_magic != XLOG_PAGE_MAGIC) { char fname[MAXFNAMELEN]; @@ -921,6 +991,14 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, LSN_FORMAT_ARGS(hdr->xlp_pageaddr), fname, offset); + + /* + * If the page address is less than expected we assume it is an unused + * page in a recycled segment. + */ + if (hdr->xlp_pageaddr < recaddr) + state->EndOfWAL = true; + return false; } diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index f9f212680b..750056acaf 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -1592,7 +1592,7 @@ PerformWalRecovery(void) /* just have to read next record after CheckPoint */ Assert(xlogreader->ReadRecPtr == CheckPointLoc); replayTLI = CheckPointTLI; - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } if (record != NULL) @@ -1706,7 +1706,7 @@ PerformWalRecovery(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } while (record != NULL); /* @@ -1765,13 +1765,20 @@ PerformWalRecovery(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that indicate @@ -2939,6 +2946,7 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); if (record == NULL) @@ -2954,6 +2962,18 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * EndRecPtr is the LSN we tried to read but failed. In the + * case of decoding error, it is at the end of the failed + * record but we don't have a means for now to know EndRecPtr + * is pointing to which of the beginning or ending of the + * failed record. + */ + ErrRecPtr = xlogreader->EndRecPtr; } if (readFile >= 0) @@ -2963,13 +2983,16 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * We only end up here without a message when XLogPageRead() failed + * in that case we already logged something, or just met end-of-WAL + * conditions. In StandbyMode that only happens if we have been + * triggered, so we shouldn't loop anymore in that case. When + * EndOfWAL is true, we don't emit that error if any immediately + * and instead will show it as a part of a decent end-of-wal + * message later. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -3000,11 +3023,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + Assert(ErrRecPtr != InvalidXLogRecPtr); + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -3017,11 +3043,16 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in pg_wal during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -3042,12 +3073,24 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * recovery ended. + * + * Emit a decent message if we met end-of-WAL. Otherwise we should + * have already emitted an error message. + */ + if (xlogreader->EndOfWAL) + ereport(LOG, + (errmsg("reached end of WAL at %X/%X on timeline %u", + LSN_FORMAT_ARGS(ErrRecPtr), replayTLI), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -3129,12 +3172,16 @@ retry: private->replayTLI, xlogreader->EndRecPtr)) { + Assert(!StandbyMode || CheckForStandbyTrigger()); + if (readFile >= 0) close(readFile); readFile = -1; readLen = 0; readSource = XLOG_FROM_ANY; + /* promotion exit is not end-of-WAL */ + xlogreader->EndOfWAL = !StandbyMode; return -1; } } @@ -3767,7 +3814,8 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + /* use currentSource as readSource is reset at failure */ + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index ceaff097b9..4f117ea4da 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -472,10 +472,9 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", - startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); + (errmsg("replication terminated by primary server at %X/%X on timeline %u.", + LSN_FORMAT_ARGS(LogstreamResult.Write), + startpointTLI))); endofwal = true; break; } diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index a6251e1a96..3745e76488 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -1176,9 +1176,16 @@ main(int argc, char **argv) exit(0); if (errormsg) - fatal_error("error in WAL record at %X/%X: %s", - LSN_FORMAT_ARGS(xlogreader_state->ReadRecPtr), - errormsg); + { + if (xlogreader_state->EndOfWAL) + pg_log_info("end of WAL at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + else + fatal_error("error in WAL record at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + } XLogReaderFree(xlogreader_state); diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 477f0efe26..7b314ef10e 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* was the last attempt EOW? */ /* * Set at the end of recovery: the start point of a partial record at the diff --git a/src/test/recovery/t/011_crash_recovery.pl b/src/test/recovery/t/011_crash_recovery.pl index 14154d1ce0..01033334d6 100644 --- a/src/test/recovery/t/011_crash_recovery.pl +++ b/src/test/recovery/t/011_crash_recovery.pl @@ -10,7 +10,9 @@ use PostgreSQL::Test::Cluster; use PostgreSQL::Test::Utils; use Test::More; use Config; +use IPC::Run; +my $reached_eow_pat = "reached end of WAL at "; my $node = PostgreSQL::Test::Cluster->new('primary'); $node->init(allows_streaming => 1); $node->start; @@ -48,7 +50,15 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), # Crash and restart the postmaster $node->stop('immediate'); +my $logstart = get_log_size($node); $node->start; +my $max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, $reached_eow_pat, $logstart)); + sleep 0.5; +} +ok ($max_attempts >= 0, "end-of-wal is logged"); # Make sure we really got a new xid cmp_ok($node->safe_psql('postgres', 'SELECT pg_current_xact_id()'), @@ -61,4 +71,100 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), $stdin .= "\\q\n"; $tx->finish; # wait for psql to quit gracefully +my $segsize = $node->safe_psql('postgres', + qq[SELECT setting FROM pg_settings WHERE name = 'wal_segment_size';]); + +# make sure no records afterwards go to the next segment +$node->safe_psql('postgres', qq[ + SELECT pg_switch_wal(); + CHECKPOINT; + CREATE TABLE t(); +]); +$node->stop('immediate'); + +# identify REDO WAL file +my $cmd = "pg_controldata -D " . $node->data_dir(); +$cmd = ['pg_controldata', '-D', $node->data_dir()]; +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stdout =~ /^Latest checkpoint's REDO WAL file:[ \t] *(.+)$/m, + "checkpoint file is identified"); +my $chkptfile = $1; + +# identify the last record +my $walfile = $node->data_dir() . "/pg_wal/$chkptfile"; +$cmd = ['pg_waldump', $walfile]; +$stdout = ''; +$stderr = ''; +my $lastrec; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +foreach my $l (split(/\r?\n/, $stdout)) +{ + $lastrec = $l; +} +ok(defined $lastrec, "last WAL record is extracted"); +ok($stderr =~ /end of WAL at ([0-9A-F\/]+): .* at \g1/, + "pg_waldump emits the correct ending message"); + +# read the last record LSN excluding leading zeroes +ok ($lastrec =~ /, lsn: 0\/0*([1-9A-F][0-9A-F]+),/, + "LSN of the last record identified"); +my $lastlsn = $1; + +# corrupt the last record +my $offset = hex($lastlsn) % $segsize; +open(my $segf, '+<', $walfile) or die "failed to open $walfile\n"; +seek($segf, $offset, 0); # halfway break the last record +print $segf "\0\0\0\0"; +close($segf); + +# pg_waldump complains about the corrupted record +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stderr =~ /fatal: error in WAL record at 0\/$lastlsn: .* at 0\/$lastlsn/, + "pg_waldump emits the correct error message"); + +# also server complains +$logstart = get_log_size($node); +$node->start; +$max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, "WARNING: invalid record length at 0/$lastlsn: wanted [0-9]+, got 0", + $logstart)); + sleep 0.5; +} +ok($max_attempts >= 0, "header error is logged at $lastlsn"); + +# no end-of-wal message should be seen this time +ok(!find_in_log($node, $reached_eow_pat, $logstart), + "false log message is not emitted"); + +$node->stop('immediate'); + done_testing(); + +#### helper routines +# return the size of logfile of $node in bytes +sub get_log_size +{ + my ($node) = @_; + + return (stat $node->logfile)[7]; +} + +# find $pat in logfile of $node after $off-th byte +sub find_in_log +{ + my ($node, $pat, $off) = @_; + + $off = 0 unless defined $off; + my $log = PostgreSQL::Test::Utils::slurp_file($node->logfile); + return 0 if (length($log) <= $off); + + $log = substr($log, $off); + + return $log =~ m/$pat/; +} -- 2.27.0 ----Next_Part(Thu_Feb_17_16_50_01_2022_997)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v16] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlogreader.c | 144 +++++++++++++++++----- src/backend/access/transam/xlogrecovery.c | 92 ++++++++++---- src/backend/replication/walreceiver.c | 7 +- src/bin/pg_waldump/pg_waldump.c | 13 +- src/include/access/xlogreader.h | 1 + src/test/recovery/t/011_crash_recovery.pl | 106 ++++++++++++++++ 6 files changed, 305 insertions(+), 58 deletions(-) diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 35029cf97d..bd0f211a23 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -42,6 +42,8 @@ static bool allocate_recordbuf(XLogReaderState *state, uint32 reclength); static int ReadPageInternal(XLogReaderState *state, XLogRecPtr pageptr, int reqLen); static void XLogReaderInvalReadState(XLogReaderState *state); +static bool ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record); static bool ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess); static bool ValidXLogRecord(XLogReaderState *state, XLogRecord *record, @@ -121,6 +123,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -292,6 +295,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); state->abortedRecPtr = InvalidXLogRecPtr; @@ -371,25 +375,21 @@ restart: Assert(pageHeaderSize <= readOff); /* - * Read the record length. + * Validate the record header. * - * NB: Even though we use an XLogRecord pointer here, the whole record - * header might not fit on this page. xl_tot_len is the first field of the - * struct, so it must be on this page (the records are MAXALIGNed), but we - * cannot access any other fields until we've verified that we got the - * whole header. - */ - record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); - total_len = record->xl_tot_len; - - /* - * If the whole record header is on this page, validate it immediately. - * Otherwise do just a basic sanity check on xl_tot_len, and validate the - * rest of the header after reading it from the next page. The xl_tot_len + * Even though we use an XLogRecord pointer here, the whole record header + * might not fit on this page. If the whole record header is on this page, + * validate it immediately. Even otherwise xl_tot_len must be on this page + * (it is the first field of MAXALIGNed records), but we still cannot + * access any further fields until we've verified that we got the whole + * header, so do just a basic sanity check on record length, and validate + * the rest of the header after reading it from the next page. The length * check is necessary here to ensure that we enter the "Need to reassemble * record" code path below; otherwise we might fail to apply * ValidXLogRecordHeader at all. */ + record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); + if (targetRecOff <= XLOG_BLCKSZ - SizeOfXLogRecord) { if (!ValidXLogRecordHeader(state, RecPtr, state->ReadRecPtr, record, @@ -399,18 +399,13 @@ restart: } else { - /* XXX: more validation should be done here */ - if (total_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: wanted %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, total_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) goto err; - } + gotheader = false; } + total_len = record->xl_tot_len; len = XLOG_BLCKSZ - RecPtr % XLOG_BLCKSZ; if (total_len > len) { @@ -588,6 +583,15 @@ err: */ state->abortedRecPtr = RecPtr; state->missingContrecPtr = targetPagePtr; + + /* + * If the message is not set yet, that means we failed to load the + * page for the record. Otherwise do not hide the existing message. + */ + if (state->errormsg_buf[0] == '\0') + report_invalid_record(state, + "missing contrecord at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); } /* @@ -719,6 +723,60 @@ XLogReaderInvalReadState(XLogReaderState *state) state->readLen = 0; } +/* + * Validate record length of an XLOG record header. + * + * This is substantially a part of ValidXLogRecordHeader. But XLogReadRecord + * needs this separate from the function in case of a partial record header. + */ +static bool +ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record) +{ + if (record->xl_tot_len == 0) + { + char *p; + char *pe; + + /* + * We are almost sure reaching the end of WAL, make sure that the + * whole page after the record is filled with zeroes. + */ + p = (char *) record; + pe = p + XLOG_BLCKSZ - (RecPtr & (XLOG_BLCKSZ - 1)); + + while (*p == 0 && p < pe) + p++; + + if (p == pe) + { + /* + * The page after the record is completely zeroed. That suggests + * we don't have a record after this point. We don't bother + * checking the pages after since they are not zeroed in the case + * of recycled segments. + */ + report_invalid_record(state, "empty record at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + return false; + } + } + + if (record->xl_tot_len < SizeOfXLogRecord) + { + report_invalid_record(state, + "invalid record length at %X/%X: wanted %u, got %u", + LSN_FORMAT_ARGS(RecPtr), + (uint32) SizeOfXLogRecord, record->xl_tot_len); + return false; + } + + return true; +} + /* * Validate an XLOG record header. * @@ -730,14 +788,9 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { - if (record->xl_tot_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: wanted %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, record->xl_tot_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) return false; - } + if (record->xl_rmid > RM_MAX_ID) { report_invalid_record(state, @@ -836,6 +889,31 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, XLogSegNoOffsetToRecPtr(segno, offset, state->segcxt.ws_segsize, recaddr); + StaticAssertStmt(XLOG_PAGE_MAGIC != 0, "XLOG_PAGE_MAGIC is zero"); + + if (hdr->xlp_magic == 0) + { + /* Regard an empty page as End-Of-WAL */ + int i; + + for (i = 0; i < XLOG_BLCKSZ && phdr[i] == 0; i++); + if (i == XLOG_BLCKSZ) + { + char fname[MAXFNAMELEN]; + + XLogFileName(fname, state->seg.ws_tli, segno, + state->segcxt.ws_segsize); + + report_invalid_record(state, + "empty page in log segment %s, offset %u", + fname, + offset); + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid magic number */ + } if (hdr->xlp_magic != XLOG_PAGE_MAGIC) { char fname[MAXFNAMELEN]; @@ -921,6 +999,14 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, LSN_FORMAT_ARGS(hdr->xlp_pageaddr), fname, offset); + + /* + * If the page address is less than expected we assume it is an unused + * page in a recycled segment. + */ + if (hdr->xlp_pageaddr < recaddr) + state->EndOfWAL = true; + return false; } diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index f9f212680b..750056acaf 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -1592,7 +1592,7 @@ PerformWalRecovery(void) /* just have to read next record after CheckPoint */ Assert(xlogreader->ReadRecPtr == CheckPointLoc); replayTLI = CheckPointTLI; - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } if (record != NULL) @@ -1706,7 +1706,7 @@ PerformWalRecovery(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } while (record != NULL); /* @@ -1765,13 +1765,20 @@ PerformWalRecovery(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that indicate @@ -2939,6 +2946,7 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); if (record == NULL) @@ -2954,6 +2962,18 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * EndRecPtr is the LSN we tried to read but failed. In the + * case of decoding error, it is at the end of the failed + * record but we don't have a means for now to know EndRecPtr + * is pointing to which of the beginning or ending of the + * failed record. + */ + ErrRecPtr = xlogreader->EndRecPtr; } if (readFile >= 0) @@ -2963,13 +2983,16 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * We only end up here without a message when XLogPageRead() failed + * in that case we already logged something, or just met end-of-WAL + * conditions. In StandbyMode that only happens if we have been + * triggered, so we shouldn't loop anymore in that case. When + * EndOfWAL is true, we don't emit that error if any immediately + * and instead will show it as a part of a decent end-of-wal + * message later. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -3000,11 +3023,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + Assert(ErrRecPtr != InvalidXLogRecPtr); + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -3017,11 +3043,16 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in pg_wal during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -3042,12 +3073,24 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * recovery ended. + * + * Emit a decent message if we met end-of-WAL. Otherwise we should + * have already emitted an error message. + */ + if (xlogreader->EndOfWAL) + ereport(LOG, + (errmsg("reached end of WAL at %X/%X on timeline %u", + LSN_FORMAT_ARGS(ErrRecPtr), replayTLI), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -3129,12 +3172,16 @@ retry: private->replayTLI, xlogreader->EndRecPtr)) { + Assert(!StandbyMode || CheckForStandbyTrigger()); + if (readFile >= 0) close(readFile); readFile = -1; readLen = 0; readSource = XLOG_FROM_ANY; + /* promotion exit is not end-of-WAL */ + xlogreader->EndOfWAL = !StandbyMode; return -1; } } @@ -3767,7 +3814,8 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + /* use currentSource as readSource is reset at failure */ + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index ceaff097b9..4f117ea4da 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -472,10 +472,9 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", - startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); + (errmsg("replication terminated by primary server at %X/%X on timeline %u.", + LSN_FORMAT_ARGS(LogstreamResult.Write), + startpointTLI))); endofwal = true; break; } diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index 2340dc247b..215abe95dc 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -1173,9 +1173,16 @@ main(int argc, char **argv) exit(0); if (errormsg) - fatal_error("error in WAL record at %X/%X: %s", - LSN_FORMAT_ARGS(xlogreader_state->ReadRecPtr), - errormsg); + { + if (xlogreader_state->EndOfWAL) + pg_log_info("end of WAL at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + else + fatal_error("error in WAL record at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + } XLogReaderFree(xlogreader_state); diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 477f0efe26..7b314ef10e 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* was the last attempt EOW? */ /* * Set at the end of recovery: the start point of a partial record at the diff --git a/src/test/recovery/t/011_crash_recovery.pl b/src/test/recovery/t/011_crash_recovery.pl index 14154d1ce0..01033334d6 100644 --- a/src/test/recovery/t/011_crash_recovery.pl +++ b/src/test/recovery/t/011_crash_recovery.pl @@ -10,7 +10,9 @@ use PostgreSQL::Test::Cluster; use PostgreSQL::Test::Utils; use Test::More; use Config; +use IPC::Run; +my $reached_eow_pat = "reached end of WAL at "; my $node = PostgreSQL::Test::Cluster->new('primary'); $node->init(allows_streaming => 1); $node->start; @@ -48,7 +50,15 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), # Crash and restart the postmaster $node->stop('immediate'); +my $logstart = get_log_size($node); $node->start; +my $max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, $reached_eow_pat, $logstart)); + sleep 0.5; +} +ok ($max_attempts >= 0, "end-of-wal is logged"); # Make sure we really got a new xid cmp_ok($node->safe_psql('postgres', 'SELECT pg_current_xact_id()'), @@ -61,4 +71,100 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), $stdin .= "\\q\n"; $tx->finish; # wait for psql to quit gracefully +my $segsize = $node->safe_psql('postgres', + qq[SELECT setting FROM pg_settings WHERE name = 'wal_segment_size';]); + +# make sure no records afterwards go to the next segment +$node->safe_psql('postgres', qq[ + SELECT pg_switch_wal(); + CHECKPOINT; + CREATE TABLE t(); +]); +$node->stop('immediate'); + +# identify REDO WAL file +my $cmd = "pg_controldata -D " . $node->data_dir(); +$cmd = ['pg_controldata', '-D', $node->data_dir()]; +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stdout =~ /^Latest checkpoint's REDO WAL file:[ \t] *(.+)$/m, + "checkpoint file is identified"); +my $chkptfile = $1; + +# identify the last record +my $walfile = $node->data_dir() . "/pg_wal/$chkptfile"; +$cmd = ['pg_waldump', $walfile]; +$stdout = ''; +$stderr = ''; +my $lastrec; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +foreach my $l (split(/\r?\n/, $stdout)) +{ + $lastrec = $l; +} +ok(defined $lastrec, "last WAL record is extracted"); +ok($stderr =~ /end of WAL at ([0-9A-F\/]+): .* at \g1/, + "pg_waldump emits the correct ending message"); + +# read the last record LSN excluding leading zeroes +ok ($lastrec =~ /, lsn: 0\/0*([1-9A-F][0-9A-F]+),/, + "LSN of the last record identified"); +my $lastlsn = $1; + +# corrupt the last record +my $offset = hex($lastlsn) % $segsize; +open(my $segf, '+<', $walfile) or die "failed to open $walfile\n"; +seek($segf, $offset, 0); # halfway break the last record +print $segf "\0\0\0\0"; +close($segf); + +# pg_waldump complains about the corrupted record +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stderr =~ /fatal: error in WAL record at 0\/$lastlsn: .* at 0\/$lastlsn/, + "pg_waldump emits the correct error message"); + +# also server complains +$logstart = get_log_size($node); +$node->start; +$max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, "WARNING: invalid record length at 0/$lastlsn: wanted [0-9]+, got 0", + $logstart)); + sleep 0.5; +} +ok($max_attempts >= 0, "header error is logged at $lastlsn"); + +# no end-of-wal message should be seen this time +ok(!find_in_log($node, $reached_eow_pat, $logstart), + "false log message is not emitted"); + +$node->stop('immediate'); + done_testing(); + +#### helper routines +# return the size of logfile of $node in bytes +sub get_log_size +{ + my ($node) = @_; + + return (stat $node->logfile)[7]; +} + +# find $pat in logfile of $node after $off-th byte +sub find_in_log +{ + my ($node, $pat, $off) = @_; + + $off = 0 unless defined $off; + my $log = PostgreSQL::Test::Utils::slurp_file($node->logfile); + return 0 if (length($log) <= $off); + + $log = substr($log, $off); + + return $log =~ m/$pat/; +} -- 2.27.0 ----Next_Part(Fri_Mar__4_09_43_59_2022_493)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v11] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlog.c | 93 +++++++++++++----- src/backend/access/transam/xlogreader.c | 78 +++++++++++++++ src/backend/replication/walreceiver.c | 7 +- src/bin/pg_waldump/pg_waldump.c | 13 ++- src/include/access/xlogreader.h | 1 + src/test/recovery/t/011_crash_recovery.pl | 110 +++++++++++++++++++++- 6 files changed, 271 insertions(+), 31 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 958220c495..618f33d342 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4480,6 +4480,7 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); if (record == NULL) @@ -4495,6 +4496,18 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * EndRecPtr is the LSN we tried to read but failed. In the + * case of decoding error, it is at the end of the failed + * record but we don't have a means for now to know EndRecPtr + * is pointing to which of the beginning or ending of the + * failed record. + */ + ErrRecPtr = xlogreader->EndRecPtr; } if (readFile >= 0) @@ -4504,13 +4517,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we get here for other than end-of-WAL, emit the error + * message right now. Otherwise the message if any is shown as a + * part of the end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4541,11 +4553,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4558,11 +4571,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4610,12 +4629,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * If we haven't emit an error message, we have safely reached the + * end-of-WAL. + */ + if (xlogreader->EndOfWAL) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(ErrRecPtr), replayTLI, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7294,7 +7334,7 @@ StartupXLOG(void) { ereport(LOG, (errmsg("database system was not properly shut down; " - "automatic recovery in progress"))); + "crash recovery in progress"))); if (recoveryTargetTLI > ControlFile->checkPointCopy.ThisTimeLineID) ereport(LOG, (errmsg("crash recovery starts in timeline %u " @@ -7544,7 +7584,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } if (record != NULL) @@ -7782,7 +7822,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } while (record != NULL); /* @@ -7842,13 +7882,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -13097,7 +13144,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 35029cf97d..03a8b42f15 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -121,6 +121,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -292,6 +293,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); state->abortedRecPtr = InvalidXLogRecPtr; @@ -588,6 +590,15 @@ err: */ state->abortedRecPtr = RecPtr; state->missingContrecPtr = targetPagePtr; + + /* + * If the message is not set yet, that means we failed to load the + * page for the record. Otherwise do not hide the existing message. + */ + if (state->errormsg_buf[0] == '\0') + report_invalid_record(state, + "missing contrecord at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); } /* @@ -730,6 +741,40 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* + * We are almost sure reaching the end of WAL, make sure that the + * whole page after the record is filled with zeroes. + */ + char *p; + char *pe; + + /* scan from the beginning of the record to the end of block */ + p = (char *) record; + pe = p + XLOG_BLCKSZ - (RecPtr & (XLOG_BLCKSZ - 1)); + + while (*p == 0 && p < pe) + p++; + + if (p == pe) + { + /* + * The page after the record is completely zeroed. That suggests + * we don't have a record after this point. We don't bother + * checking the pages after since they are not zeroed in the case + * of recycled segments. + */ + report_invalid_record(state, "empty record found at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid record length */ + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, @@ -836,6 +881,31 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, XLogSegNoOffsetToRecPtr(segno, offset, state->segcxt.ws_segsize, recaddr); + StaticAssertStmt(XLOG_PAGE_MAGIC != 0, "XLOG_PAGE_MAGIC is zero"); + + if (hdr->xlp_magic == 0) + { + /* Regard an empty page as End-Of-WAL */ + int i; + + for (i = 0; i < XLOG_BLCKSZ && phdr[i] == 0; i++); + if (i == XLOG_BLCKSZ) + { + char fname[MAXFNAMELEN]; + + XLogFileName(fname, state->seg.ws_tli, segno, + state->segcxt.ws_segsize); + + report_invalid_record(state, + "empty page in log segment %s, offset %u", + fname, + offset); + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid magic number */ + } if (hdr->xlp_magic != XLOG_PAGE_MAGIC) { char fname[MAXFNAMELEN]; @@ -921,6 +991,14 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, LSN_FORMAT_ARGS(hdr->xlp_pageaddr), fname, offset); + + /* + * If the page address is less than expected we assume it is an unused + * page in a recycled segment. + */ + if (hdr->xlp_pageaddr < recaddr) + state->EndOfWAL = true; + return false; } diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index b39fce8c23..1a7a692bc0 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -471,10 +471,9 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", - startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); + (errmsg("replication terminated by primary server at %X/%X on timeline %u.", + LSN_FORMAT_ARGS(LogstreamResult.Write), + startpointTLI))); endofwal = true; break; } diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index a6251e1a96..3745e76488 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -1176,9 +1176,16 @@ main(int argc, char **argv) exit(0); if (errormsg) - fatal_error("error in WAL record at %X/%X: %s", - LSN_FORMAT_ARGS(xlogreader_state->ReadRecPtr), - errormsg); + { + if (xlogreader_state->EndOfWAL) + pg_log_info("end of WAL at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + else + fatal_error("error in WAL record at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + } XLogReaderFree(xlogreader_state); diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 477f0efe26..3eeba220a1 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* * Set at the end of recovery: the start point of a partial record at the diff --git a/src/test/recovery/t/011_crash_recovery.pl b/src/test/recovery/t/011_crash_recovery.pl index 3892aba3e5..67d264df26 100644 --- a/src/test/recovery/t/011_crash_recovery.pl +++ b/src/test/recovery/t/011_crash_recovery.pl @@ -10,9 +10,11 @@ use PostgreSQL::Test::Cluster; use PostgreSQL::Test::Utils; use Test::More; use Config; +use IPC::Run; -plan tests => 3; +plan tests => 11; +my $reached_eow_pat = "reached end of WAL at "; my $node = PostgreSQL::Test::Cluster->new('primary'); $node->init(allows_streaming => 1); $node->start; @@ -50,7 +52,15 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), # Crash and restart the postmaster $node->stop('immediate'); +my $logstart = get_log_size($node); $node->start; +my $max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, $reached_eow_pat, $logstart)); + sleep 0.5; +} +ok ($max_attempts >= 0, "end-of-wal is logged"); # Make sure we really got a new xid cmp_ok($node->safe_psql('postgres', 'SELECT pg_current_xact_id()'), @@ -62,3 +72,101 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), $stdin .= "\\q\n"; $tx->finish; # wait for psql to quit gracefully + +my $segsize = $node->safe_psql('postgres', + qq[SELECT setting FROM pg_settings WHERE name = 'wal_segment_size';]); + +# make sure no records afterwards go to the next segment +$node->safe_psql('postgres', qq[ + SELECT pg_switch_wal(); + CHECKPOINT; + CREATE TABLE t(); +]); +$node->stop('immediate'); + +# identify REDO WAL file +my $cmd = "pg_controldata -D " . $node->data_dir(); +my $chkptfile; +$cmd = ['pg_controldata', '-D', $node->data_dir()]; +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stdout =~ /^Latest checkpoint's REDO WAL file:[ \t] *(.+)$/m, + "checkpoint file is identified"); +my $chkptfile = $1; + +# identify the last record +my $walfile = $node->data_dir() . "/pg_wal/$chkptfile"; +$cmd = ['pg_waldump', $walfile]; +$stdout = ''; +$stderr = ''; +my $lastrec; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +foreach my $l (split(/\r?\n/, $stdout)) +{ + $lastrec = $l; +} +ok(defined $lastrec, "last WAL record is extracted"); +ok($stderr =~ /end of WAL at ([0-9A-F\/]+): .* at \g1/, + "pg_waldump emits the correct ending message"); + +# read the last record LSN excluding leading zeroes +ok ($lastrec =~ /, lsn: 0\/0*([1-9A-F][0-9A-F]+),/, + "LSN of the last record identified"); +my $lastlsn = $1; + +# corrupt the last record +my $offset = hex($lastlsn) % $segsize; +open(my $segf, '+<', $walfile) or die "failed to open $walfile\n"; +seek($segf, $offset, 0); # halfway break the last record +print $segf "\0\0\0\0"; +close($segf); + +# pg_waldump complains about the corrupted record +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stderr =~ /fatal: error in WAL record at 0\/$lastlsn: .* at 0\/$lastlsn/, + "pg_waldump emits the correct error message"); + +# also server complains +$logstart = get_log_size($node); +$node->start; +$max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, "WARNING: invalid record length at 0/$lastlsn: wanted [0-9]+, got 0", + $logstart)); + sleep 0.5; +} +ok($max_attempts >= 0, "header error is logged at $lastlsn"); + +# and the end-of-wal messages shouldn't be seen +# the same message has been confirmed in the past +ok(!find_in_log($node, $reached_eow_pat, $logstart), + "false log message is not emitted"); + +$node->stop('immediate'); + +#### helper routines +# return the size of logfile of $node in bytes +sub get_log_size +{ + my ($node) = @_; + + return (stat $node->logfile)[7]; +} + +# find $pat in logfile of $node after $off-th byte +sub find_in_log +{ + my ($node, $pat, $off) = @_; + + $off = 0 unless defined $off; + my $log = PostgreSQL::Test::Utils::slurp_file($node->logfile); + return 0 if (length($log) <= $off); + + $log = substr($log, $off); + + return $log =~ m/$pat/; +} -- 2.27.0 ----Next_Part(Wed_Feb__9_16_44_14_2022_277)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v12] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlog.c | 92 +++++++++++++----- src/backend/access/transam/xlogreader.c | 78 ++++++++++++++++ src/backend/replication/walreceiver.c | 7 +- src/bin/pg_waldump/pg_waldump.c | 13 ++- src/include/access/xlogreader.h | 1 + src/test/recovery/t/011_crash_recovery.pl | 108 +++++++++++++++++++++- 6 files changed, 268 insertions(+), 31 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 958220c495..bf1d40e7cb 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4480,6 +4480,7 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); if (record == NULL) @@ -4495,6 +4496,18 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * EndRecPtr is the LSN we tried to read but failed. In the + * case of decoding error, it is at the end of the failed + * record but we don't have a means for now to know EndRecPtr + * is pointing to which of the beginning or ending of the + * failed record. + */ + ErrRecPtr = xlogreader->EndRecPtr; } if (readFile >= 0) @@ -4504,13 +4517,16 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * We only end up here without a message when XLogPageRead() failed + * in that case we already logged something, or just met end-of-WAL + * conditions. In StandbyMode that only happens if we have been + * triggered, so we shouldn't loop anymore in that case. When + * EndOfWAL is true, we don't emit that error if any immediately + * and instead will show it as a part of a decent end-of-wal + * message later. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4541,11 +4557,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + Assert(ErrRecPtr != InvalidXLogRecPtr); + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4558,11 +4577,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4610,12 +4635,24 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * recovery ended. + * + * Emit a decent message if we met end-of-WAL. Otherwise we should + * have already emitted an error message. + */ + if (xlogreader->EndOfWAL) + ereport(LOG, + (errmsg("reached end of WAL at %X/%X on timeline %u", + LSN_FORMAT_ARGS(ErrRecPtr), replayTLI), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7544,7 +7581,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } if (record != NULL) @@ -7782,7 +7819,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } while (record != NULL); /* @@ -7842,13 +7879,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12434,12 +12478,14 @@ retry: private->replayTLI, xlogreader->EndRecPtr)) { + Assert(!StandbyMode); + if (readFile >= 0) close(readFile); readFile = -1; readLen = 0; readSource = XLOG_FROM_ANY; - + xlogreader->EndOfWAL = true; return -1; } } @@ -13097,7 +13143,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 35029cf97d..22982c4de7 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -121,6 +121,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -292,6 +293,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); state->abortedRecPtr = InvalidXLogRecPtr; @@ -588,6 +590,15 @@ err: */ state->abortedRecPtr = RecPtr; state->missingContrecPtr = targetPagePtr; + + /* + * If the message is not set yet, that means we failed to load the + * page for the record. Otherwise do not hide the existing message. + */ + if (state->errormsg_buf[0] == '\0') + report_invalid_record(state, + "missing contrecord at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); } /* @@ -730,6 +741,40 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* + * We are almost sure reaching the end of WAL, make sure that the + * whole page after the record is filled with zeroes. + */ + char *p; + char *pe; + + /* scan from the beginning of the record to the end of block */ + p = (char *) record; + pe = p + XLOG_BLCKSZ - (RecPtr & (XLOG_BLCKSZ - 1)); + + while (*p == 0 && p < pe) + p++; + + if (p == pe) + { + /* + * The page after the record is completely zeroed. That suggests + * we don't have a record after this point. We don't bother + * checking the pages after since they are not zeroed in the case + * of recycled segments. + */ + report_invalid_record(state, "empty record at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid record length */ + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, @@ -836,6 +881,31 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, XLogSegNoOffsetToRecPtr(segno, offset, state->segcxt.ws_segsize, recaddr); + StaticAssertStmt(XLOG_PAGE_MAGIC != 0, "XLOG_PAGE_MAGIC is zero"); + + if (hdr->xlp_magic == 0) + { + /* Regard an empty page as End-Of-WAL */ + int i; + + for (i = 0; i < XLOG_BLCKSZ && phdr[i] == 0; i++); + if (i == XLOG_BLCKSZ) + { + char fname[MAXFNAMELEN]; + + XLogFileName(fname, state->seg.ws_tli, segno, + state->segcxt.ws_segsize); + + report_invalid_record(state, + "empty page in log segment %s, offset %u", + fname, + offset); + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid magic number */ + } if (hdr->xlp_magic != XLOG_PAGE_MAGIC) { char fname[MAXFNAMELEN]; @@ -921,6 +991,14 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, LSN_FORMAT_ARGS(hdr->xlp_pageaddr), fname, offset); + + /* + * If the page address is less than expected we assume it is an unused + * page in a recycled segment. + */ + if (hdr->xlp_pageaddr < recaddr) + state->EndOfWAL = true; + return false; } diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index b39fce8c23..1a7a692bc0 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -471,10 +471,9 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", - startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); + (errmsg("replication terminated by primary server at %X/%X on timeline %u.", + LSN_FORMAT_ARGS(LogstreamResult.Write), + startpointTLI))); endofwal = true; break; } diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index a6251e1a96..3745e76488 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -1176,9 +1176,16 @@ main(int argc, char **argv) exit(0); if (errormsg) - fatal_error("error in WAL record at %X/%X: %s", - LSN_FORMAT_ARGS(xlogreader_state->ReadRecPtr), - errormsg); + { + if (xlogreader_state->EndOfWAL) + pg_log_info("end of WAL at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + else + fatal_error("error in WAL record at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + } XLogReaderFree(xlogreader_state); diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 477f0efe26..7b314ef10e 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* was the last attempt EOW? */ /* * Set at the end of recovery: the start point of a partial record at the diff --git a/src/test/recovery/t/011_crash_recovery.pl b/src/test/recovery/t/011_crash_recovery.pl index 3892aba3e5..1d7476c309 100644 --- a/src/test/recovery/t/011_crash_recovery.pl +++ b/src/test/recovery/t/011_crash_recovery.pl @@ -10,9 +10,11 @@ use PostgreSQL::Test::Cluster; use PostgreSQL::Test::Utils; use Test::More; use Config; +use IPC::Run; -plan tests => 3; +plan tests => 11; +my $reached_eow_pat = "reached end of WAL at "; my $node = PostgreSQL::Test::Cluster->new('primary'); $node->init(allows_streaming => 1); $node->start; @@ -50,7 +52,15 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), # Crash and restart the postmaster $node->stop('immediate'); +my $logstart = get_log_size($node); $node->start; +my $max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, $reached_eow_pat, $logstart)); + sleep 0.5; +} +ok ($max_attempts >= 0, "end-of-wal is logged"); # Make sure we really got a new xid cmp_ok($node->safe_psql('postgres', 'SELECT pg_current_xact_id()'), @@ -62,3 +72,99 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), $stdin .= "\\q\n"; $tx->finish; # wait for psql to quit gracefully + +my $segsize = $node->safe_psql('postgres', + qq[SELECT setting FROM pg_settings WHERE name = 'wal_segment_size';]); + +# make sure no records afterwards go to the next segment +$node->safe_psql('postgres', qq[ + SELECT pg_switch_wal(); + CHECKPOINT; + CREATE TABLE t(); +]); +$node->stop('immediate'); + +# identify REDO WAL file +my $cmd = "pg_controldata -D " . $node->data_dir(); +$cmd = ['pg_controldata', '-D', $node->data_dir()]; +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stdout =~ /^Latest checkpoint's REDO WAL file:[ \t] *(.+)$/m, + "checkpoint file is identified"); +my $chkptfile = $1; + +# identify the last record +my $walfile = $node->data_dir() . "/pg_wal/$chkptfile"; +$cmd = ['pg_waldump', $walfile]; +$stdout = ''; +$stderr = ''; +my $lastrec; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +foreach my $l (split(/\r?\n/, $stdout)) +{ + $lastrec = $l; +} +ok(defined $lastrec, "last WAL record is extracted"); +ok($stderr =~ /end of WAL at ([0-9A-F\/]+): .* at \g1/, + "pg_waldump emits the correct ending message"); + +# read the last record LSN excluding leading zeroes +ok ($lastrec =~ /, lsn: 0\/0*([1-9A-F][0-9A-F]+),/, + "LSN of the last record identified"); +my $lastlsn = $1; + +# corrupt the last record +my $offset = hex($lastlsn) % $segsize; +open(my $segf, '+<', $walfile) or die "failed to open $walfile\n"; +seek($segf, $offset, 0); # halfway break the last record +print $segf "\0\0\0\0"; +close($segf); + +# pg_waldump complains about the corrupted record +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stderr =~ /fatal: error in WAL record at 0\/$lastlsn: .* at 0\/$lastlsn/, + "pg_waldump emits the correct error message"); + +# also server complains +$logstart = get_log_size($node); +$node->start; +$max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, "WARNING: invalid record length at 0/$lastlsn: wanted [0-9]+, got 0", + $logstart)); + sleep 0.5; +} +ok($max_attempts >= 0, "header error is logged at $lastlsn"); + +# no end-of-wal message should be seen this time +ok(!find_in_log($node, $reached_eow_pat, $logstart), + "false log message is not emitted"); + +$node->stop('immediate'); + +#### helper routines +# return the size of logfile of $node in bytes +sub get_log_size +{ + my ($node) = @_; + + return (stat $node->logfile)[7]; +} + +# find $pat in logfile of $node after $off-th byte +sub find_in_log +{ + my ($node, $pat, $off) = @_; + + $off = 0 unless defined $off; + my $log = PostgreSQL::Test::Utils::slurp_file($node->logfile); + return 0 if (length($log) <= $off); + + $log = substr($log, $off); + + return $log =~ m/$pat/; +} -- 2.27.0 ----Next_Part(Thu_Feb_10_15_17_36_2022_468)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v13] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlog.c | 91 ++++++++++++++----- src/backend/access/transam/xlogreader.c | 78 ++++++++++++++++ src/backend/replication/walreceiver.c | 7 +- src/bin/pg_waldump/pg_waldump.c | 13 ++- src/include/access/xlogreader.h | 1 + src/test/recovery/t/011_crash_recovery.pl | 106 ++++++++++++++++++++++ 6 files changed, 266 insertions(+), 30 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 958220c495..bb7026ac77 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4480,6 +4480,7 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); if (record == NULL) @@ -4495,6 +4496,18 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * EndRecPtr is the LSN we tried to read but failed. In the + * case of decoding error, it is at the end of the failed + * record but we don't have a means for now to know EndRecPtr + * is pointing to which of the beginning or ending of the + * failed record. + */ + ErrRecPtr = xlogreader->EndRecPtr; } if (readFile >= 0) @@ -4504,13 +4517,16 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * We only end up here without a message when XLogPageRead() failed + * in that case we already logged something, or just met end-of-WAL + * conditions. In StandbyMode that only happens if we have been + * triggered, so we shouldn't loop anymore in that case. When + * EndOfWAL is true, we don't emit that error if any immediately + * and instead will show it as a part of a decent end-of-wal + * message later. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4541,11 +4557,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + Assert(ErrRecPtr != InvalidXLogRecPtr); + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4558,11 +4577,16 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in pg_wal during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4610,12 +4634,24 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * recovery ended. + * + * Emit a decent message if we met end-of-WAL. Otherwise we should + * have already emitted an error message. + */ + if (xlogreader->EndOfWAL) + ereport(LOG, + (errmsg("reached end of WAL at %X/%X on timeline %u", + LSN_FORMAT_ARGS(ErrRecPtr), replayTLI), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7544,7 +7580,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } if (record != NULL) @@ -7782,7 +7818,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } while (record != NULL); /* @@ -7842,13 +7878,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12434,12 +12477,14 @@ retry: private->replayTLI, xlogreader->EndRecPtr)) { + Assert(!StandbyMode); + if (readFile >= 0) close(readFile); readFile = -1; readLen = 0; readSource = XLOG_FROM_ANY; - + xlogreader->EndOfWAL = true; return -1; } } @@ -13097,7 +13142,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 35029cf97d..22982c4de7 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -121,6 +121,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -292,6 +293,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); state->abortedRecPtr = InvalidXLogRecPtr; @@ -588,6 +590,15 @@ err: */ state->abortedRecPtr = RecPtr; state->missingContrecPtr = targetPagePtr; + + /* + * If the message is not set yet, that means we failed to load the + * page for the record. Otherwise do not hide the existing message. + */ + if (state->errormsg_buf[0] == '\0') + report_invalid_record(state, + "missing contrecord at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); } /* @@ -730,6 +741,40 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* + * We are almost sure reaching the end of WAL, make sure that the + * whole page after the record is filled with zeroes. + */ + char *p; + char *pe; + + /* scan from the beginning of the record to the end of block */ + p = (char *) record; + pe = p + XLOG_BLCKSZ - (RecPtr & (XLOG_BLCKSZ - 1)); + + while (*p == 0 && p < pe) + p++; + + if (p == pe) + { + /* + * The page after the record is completely zeroed. That suggests + * we don't have a record after this point. We don't bother + * checking the pages after since they are not zeroed in the case + * of recycled segments. + */ + report_invalid_record(state, "empty record at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid record length */ + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, @@ -836,6 +881,31 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, XLogSegNoOffsetToRecPtr(segno, offset, state->segcxt.ws_segsize, recaddr); + StaticAssertStmt(XLOG_PAGE_MAGIC != 0, "XLOG_PAGE_MAGIC is zero"); + + if (hdr->xlp_magic == 0) + { + /* Regard an empty page as End-Of-WAL */ + int i; + + for (i = 0; i < XLOG_BLCKSZ && phdr[i] == 0; i++); + if (i == XLOG_BLCKSZ) + { + char fname[MAXFNAMELEN]; + + XLogFileName(fname, state->seg.ws_tli, segno, + state->segcxt.ws_segsize); + + report_invalid_record(state, + "empty page in log segment %s, offset %u", + fname, + offset); + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid magic number */ + } if (hdr->xlp_magic != XLOG_PAGE_MAGIC) { char fname[MAXFNAMELEN]; @@ -921,6 +991,14 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, LSN_FORMAT_ARGS(hdr->xlp_pageaddr), fname, offset); + + /* + * If the page address is less than expected we assume it is an unused + * page in a recycled segment. + */ + if (hdr->xlp_pageaddr < recaddr) + state->EndOfWAL = true; + return false; } diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index b39fce8c23..1a7a692bc0 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -471,10 +471,9 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", - startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); + (errmsg("replication terminated by primary server at %X/%X on timeline %u.", + LSN_FORMAT_ARGS(LogstreamResult.Write), + startpointTLI))); endofwal = true; break; } diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index a6251e1a96..3745e76488 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -1176,9 +1176,16 @@ main(int argc, char **argv) exit(0); if (errormsg) - fatal_error("error in WAL record at %X/%X: %s", - LSN_FORMAT_ARGS(xlogreader_state->ReadRecPtr), - errormsg); + { + if (xlogreader_state->EndOfWAL) + pg_log_info("end of WAL at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + else + fatal_error("error in WAL record at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + } XLogReaderFree(xlogreader_state); diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 477f0efe26..7b314ef10e 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* was the last attempt EOW? */ /* * Set at the end of recovery: the start point of a partial record at the diff --git a/src/test/recovery/t/011_crash_recovery.pl b/src/test/recovery/t/011_crash_recovery.pl index 14154d1ce0..01033334d6 100644 --- a/src/test/recovery/t/011_crash_recovery.pl +++ b/src/test/recovery/t/011_crash_recovery.pl @@ -10,7 +10,9 @@ use PostgreSQL::Test::Cluster; use PostgreSQL::Test::Utils; use Test::More; use Config; +use IPC::Run; +my $reached_eow_pat = "reached end of WAL at "; my $node = PostgreSQL::Test::Cluster->new('primary'); $node->init(allows_streaming => 1); $node->start; @@ -48,7 +50,15 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), # Crash and restart the postmaster $node->stop('immediate'); +my $logstart = get_log_size($node); $node->start; +my $max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, $reached_eow_pat, $logstart)); + sleep 0.5; +} +ok ($max_attempts >= 0, "end-of-wal is logged"); # Make sure we really got a new xid cmp_ok($node->safe_psql('postgres', 'SELECT pg_current_xact_id()'), @@ -61,4 +71,100 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), $stdin .= "\\q\n"; $tx->finish; # wait for psql to quit gracefully +my $segsize = $node->safe_psql('postgres', + qq[SELECT setting FROM pg_settings WHERE name = 'wal_segment_size';]); + +# make sure no records afterwards go to the next segment +$node->safe_psql('postgres', qq[ + SELECT pg_switch_wal(); + CHECKPOINT; + CREATE TABLE t(); +]); +$node->stop('immediate'); + +# identify REDO WAL file +my $cmd = "pg_controldata -D " . $node->data_dir(); +$cmd = ['pg_controldata', '-D', $node->data_dir()]; +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stdout =~ /^Latest checkpoint's REDO WAL file:[ \t] *(.+)$/m, + "checkpoint file is identified"); +my $chkptfile = $1; + +# identify the last record +my $walfile = $node->data_dir() . "/pg_wal/$chkptfile"; +$cmd = ['pg_waldump', $walfile]; +$stdout = ''; +$stderr = ''; +my $lastrec; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +foreach my $l (split(/\r?\n/, $stdout)) +{ + $lastrec = $l; +} +ok(defined $lastrec, "last WAL record is extracted"); +ok($stderr =~ /end of WAL at ([0-9A-F\/]+): .* at \g1/, + "pg_waldump emits the correct ending message"); + +# read the last record LSN excluding leading zeroes +ok ($lastrec =~ /, lsn: 0\/0*([1-9A-F][0-9A-F]+),/, + "LSN of the last record identified"); +my $lastlsn = $1; + +# corrupt the last record +my $offset = hex($lastlsn) % $segsize; +open(my $segf, '+<', $walfile) or die "failed to open $walfile\n"; +seek($segf, $offset, 0); # halfway break the last record +print $segf "\0\0\0\0"; +close($segf); + +# pg_waldump complains about the corrupted record +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stderr =~ /fatal: error in WAL record at 0\/$lastlsn: .* at 0\/$lastlsn/, + "pg_waldump emits the correct error message"); + +# also server complains +$logstart = get_log_size($node); +$node->start; +$max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, "WARNING: invalid record length at 0/$lastlsn: wanted [0-9]+, got 0", + $logstart)); + sleep 0.5; +} +ok($max_attempts >= 0, "header error is logged at $lastlsn"); + +# no end-of-wal message should be seen this time +ok(!find_in_log($node, $reached_eow_pat, $logstart), + "false log message is not emitted"); + +$node->stop('immediate'); + done_testing(); + +#### helper routines +# return the size of logfile of $node in bytes +sub get_log_size +{ + my ($node) = @_; + + return (stat $node->logfile)[7]; +} + +# find $pat in logfile of $node after $off-th byte +sub find_in_log +{ + my ($node, $pat, $off) = @_; + + $off = 0 unless defined $off; + my $log = PostgreSQL::Test::Utils::slurp_file($node->logfile); + return 0 if (length($log) <= $off); + + $log = substr($log, $off); + + return $log =~ m/$pat/; +} -- 2.27.0 ----Next_Part(Tue_Feb_15_11_22_38_2022_332)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v18] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlogreader.c | 145 +++++++++++++++++----- src/backend/access/transam/xlogrecovery.c | 92 ++++++++++---- src/backend/replication/walreceiver.c | 7 +- src/bin/pg_waldump/pg_waldump.c | 13 +- src/include/access/xlogreader.h | 1 + src/test/recovery/t/011_crash_recovery.pl | 106 ++++++++++++++++ 6 files changed, 306 insertions(+), 58 deletions(-) diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index e437c42992..0942265408 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -46,6 +46,8 @@ static int ReadPageInternal(XLogReaderState *state, XLogRecPtr pageptr, int reqLen); static void XLogReaderInvalReadState(XLogReaderState *state); static XLogPageReadResult XLogDecodeNextRecord(XLogReaderState *state, bool non_blocking); +static bool ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record); static bool ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess); static bool ValidXLogRecord(XLogReaderState *state, XLogRecord *record, @@ -147,6 +149,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -552,6 +555,7 @@ XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking) /* reset error state */ state->errormsg_buf[0] = '\0'; decoded = NULL; + state->EndOfWAL = false; state->abortedRecPtr = InvalidXLogRecPtr; state->missingContrecPtr = InvalidXLogRecPtr; @@ -633,25 +637,21 @@ restart: Assert(pageHeaderSize <= readOff); /* - * Read the record length. + * Validate the record header. * - * NB: Even though we use an XLogRecord pointer here, the whole record - * header might not fit on this page. xl_tot_len is the first field of the - * struct, so it must be on this page (the records are MAXALIGNed), but we - * cannot access any other fields until we've verified that we got the - * whole header. - */ - record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); - total_len = record->xl_tot_len; - - /* - * If the whole record header is on this page, validate it immediately. - * Otherwise do just a basic sanity check on xl_tot_len, and validate the - * rest of the header after reading it from the next page. The xl_tot_len + * Even though we use an XLogRecord pointer here, the whole record header + * might not fit on this page. If the whole record header is on this page, + * validate it immediately. Even otherwise xl_tot_len must be on this page + * (it is the first field of MAXALIGNed records), but we still cannot + * access any further fields until we've verified that we got the whole + * header, so do just a basic sanity check on record length, and validate + * the rest of the header after reading it from the next page. The length * check is necessary here to ensure that we enter the "Need to reassemble * record" code path below; otherwise we might fail to apply * ValidXLogRecordHeader at all. */ + record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); + if (targetRecOff <= XLOG_BLCKSZ - SizeOfXLogRecord) { if (!ValidXLogRecordHeader(state, RecPtr, state->DecodeRecPtr, record, @@ -661,18 +661,14 @@ restart: } else { - /* XXX: more validation should be done here */ - if (total_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: wanted %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, total_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) goto err; - } + gotheader = false; } + total_len = record->xl_tot_len; + /* * Find space to decode this record. Don't allow oversized allocation if * the caller requested nonblocking. Otherwise, we *have* to try to @@ -904,6 +900,15 @@ err: */ state->abortedRecPtr = RecPtr; state->missingContrecPtr = targetPagePtr; + + /* + * If the message is not set yet, that means we failed to load the + * page for the record. Otherwise do not hide the existing message. + */ + if (state->errormsg_buf[0] == '\0') + report_invalid_record(state, + "missing contrecord at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); } if (decoded && decoded->oversized) @@ -1083,6 +1088,60 @@ XLogReaderInvalReadState(XLogReaderState *state) state->readLen = 0; } +/* + * Validate record length of an XLOG record header. + * + * This is substantially a part of ValidXLogRecordHeader. But XLogReadRecord + * needs this separate from the function in case of a partial record header. + */ +static bool +ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record) +{ + if (record->xl_tot_len == 0) + { + char *p; + char *pe; + + /* + * We are almost sure reaching the end of WAL, make sure that the + * whole page after the record is filled with zeroes. + */ + p = (char *) record; + pe = p + XLOG_BLCKSZ - (RecPtr & (XLOG_BLCKSZ - 1)); + + while (*p == 0 && p < pe) + p++; + + if (p == pe) + { + /* + * The page after the record is completely zeroed. That suggests + * we don't have a record after this point. We don't bother + * checking the pages after since they are not zeroed in the case + * of recycled segments. + */ + report_invalid_record(state, "empty record at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + return false; + } + } + + if (record->xl_tot_len < SizeOfXLogRecord) + { + report_invalid_record(state, + "invalid record length at %X/%X: wanted %u, got %u", + LSN_FORMAT_ARGS(RecPtr), + (uint32) SizeOfXLogRecord, record->xl_tot_len); + return false; + } + + return true; +} + /* * Validate an XLOG record header. * @@ -1094,14 +1153,9 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { - if (record->xl_tot_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: wanted %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, record->xl_tot_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) return false; - } + if (record->xl_rmid > RM_MAX_ID) { report_invalid_record(state, @@ -1200,6 +1254,31 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, XLogSegNoOffsetToRecPtr(segno, offset, state->segcxt.ws_segsize, recaddr); + StaticAssertStmt(XLOG_PAGE_MAGIC != 0, "XLOG_PAGE_MAGIC is zero"); + + if (hdr->xlp_magic == 0) + { + /* Regard an empty page as End-Of-WAL */ + int i; + + for (i = 0; i < XLOG_BLCKSZ && phdr[i] == 0; i++); + if (i == XLOG_BLCKSZ) + { + char fname[MAXFNAMELEN]; + + XLogFileName(fname, state->seg.ws_tli, segno, + state->segcxt.ws_segsize); + + report_invalid_record(state, + "empty page in log segment %s, offset %u", + fname, + offset); + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid magic number */ + } if (hdr->xlp_magic != XLOG_PAGE_MAGIC) { char fname[MAXFNAMELEN]; @@ -1285,6 +1364,14 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, LSN_FORMAT_ARGS(hdr->xlp_pageaddr), fname, offset); + + /* + * If the page address is less than expected we assume it is an unused + * page in a recycled segment. + */ + if (hdr->xlp_pageaddr < recaddr) + state->EndOfWAL = true; + return false; } diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index 8b22c4e634..de8be3b834 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -1592,7 +1592,7 @@ PerformWalRecovery(void) /* just have to read next record after CheckPoint */ Assert(xlogreader->ReadRecPtr == CheckPointLoc); replayTLI = CheckPointTLI; - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } if (record != NULL) @@ -1706,7 +1706,7 @@ PerformWalRecovery(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } while (record != NULL); /* @@ -1765,13 +1765,20 @@ PerformWalRecovery(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that indicate @@ -2949,6 +2956,7 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); if (record == NULL) @@ -2964,6 +2972,18 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * EndRecPtr is the LSN we tried to read but failed. In the + * case of decoding error, it is at the end of the failed + * record but we don't have a means for now to know EndRecPtr + * is pointing to which of the beginning or ending of the + * failed record. + */ + ErrRecPtr = xlogreader->EndRecPtr; } if (readFile >= 0) @@ -2973,13 +2993,16 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * We only end up here without a message when XLogPageRead() failed + * in that case we already logged something, or just met end-of-WAL + * conditions. In StandbyMode that only happens if we have been + * triggered, so we shouldn't loop anymore in that case. When + * EndOfWAL is true, we don't emit that error if any immediately + * and instead will show it as a part of a decent end-of-wal + * message later. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -3010,11 +3033,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + Assert(ErrRecPtr != InvalidXLogRecPtr); + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -3027,11 +3053,16 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in pg_wal during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -3052,12 +3083,24 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * recovery ended. + * + * Emit a decent message if we met end-of-WAL. Otherwise we should + * have already emitted an error message. + */ + if (xlogreader->EndOfWAL) + ereport(LOG, + (errmsg("reached end of WAL at %X/%X on timeline %u", + LSN_FORMAT_ARGS(ErrRecPtr), replayTLI), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -3139,12 +3182,16 @@ retry: private->replayTLI, xlogreader->EndRecPtr)) { + Assert(!StandbyMode || CheckForStandbyTrigger()); + if (readFile >= 0) close(readFile); readFile = -1; readLen = 0; readSource = XLOG_FROM_ANY; + /* promotion exit is not end-of-WAL */ + xlogreader->EndOfWAL = !StandbyMode; return -1; } } @@ -3777,7 +3824,8 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + /* use currentSource as readSource is reset at failure */ + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index 3c9411e221..2f9ef9bf31 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -472,10 +472,9 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", - startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); + (errmsg("replication terminated by primary server at %X/%X on timeline %u.", + LSN_FORMAT_ARGS(LogstreamResult.Write), + startpointTLI))); endofwal = true; break; } diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index 4cb40d068a..2a78e954de 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -1310,9 +1310,16 @@ main(int argc, char **argv) exit(0); if (errormsg) - fatal_error("error in WAL record at %X/%X: %s", - LSN_FORMAT_ARGS(xlogreader_state->ReadRecPtr), - errormsg); + { + if (xlogreader_state->EndOfWAL) + pg_log_info("end of WAL at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + else + fatal_error("error in WAL record at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + } XLogReaderFree(xlogreader_state); diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index f4388cc9be..21a8f9552c 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -201,6 +201,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* was the last attempt EOW? */ /* * Set at the end of recovery: the start point of a partial record at the diff --git a/src/test/recovery/t/011_crash_recovery.pl b/src/test/recovery/t/011_crash_recovery.pl index 1b57d01046..f8b4a8417c 100644 --- a/src/test/recovery/t/011_crash_recovery.pl +++ b/src/test/recovery/t/011_crash_recovery.pl @@ -9,7 +9,9 @@ use warnings; use PostgreSQL::Test::Cluster; use PostgreSQL::Test::Utils; use Test::More; +use IPC::Run; +my $reached_eow_pat = "reached end of WAL at "; my $node = PostgreSQL::Test::Cluster->new('primary'); $node->init(allows_streaming => 1); $node->start; @@ -47,7 +49,15 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), # Crash and restart the postmaster $node->stop('immediate'); +my $logstart = get_log_size($node); $node->start; +my $max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, $reached_eow_pat, $logstart)); + sleep 0.5; +} +ok ($max_attempts >= 0, "end-of-wal is logged"); # Make sure we really got a new xid cmp_ok($node->safe_psql('postgres', 'SELECT pg_current_xact_id()'), @@ -60,4 +70,100 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), $stdin .= "\\q\n"; $tx->finish; # wait for psql to quit gracefully +my $segsize = $node->safe_psql('postgres', + qq[SELECT setting FROM pg_settings WHERE name = 'wal_segment_size';]); + +# make sure no records afterwards go to the next segment +$node->safe_psql('postgres', qq[ + SELECT pg_switch_wal(); + CHECKPOINT; + CREATE TABLE t(); +]); +$node->stop('immediate'); + +# identify REDO WAL file +my $cmd = "pg_controldata -D " . $node->data_dir(); +$cmd = ['pg_controldata', '-D', $node->data_dir()]; +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stdout =~ /^Latest checkpoint's REDO WAL file:[ \t] *(.+)$/m, + "checkpoint file is identified"); +my $chkptfile = $1; + +# identify the last record +my $walfile = $node->data_dir() . "/pg_wal/$chkptfile"; +$cmd = ['pg_waldump', $walfile]; +$stdout = ''; +$stderr = ''; +my $lastrec; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +foreach my $l (split(/\r?\n/, $stdout)) +{ + $lastrec = $l; +} +ok(defined $lastrec, "last WAL record is extracted"); +ok($stderr =~ /end of WAL at ([0-9A-F\/]+): .* at \g1/, + "pg_waldump emits the correct ending message"); + +# read the last record LSN excluding leading zeroes +ok ($lastrec =~ /, lsn: 0\/0*([1-9A-F][0-9A-F]+),/, + "LSN of the last record identified"); +my $lastlsn = $1; + +# corrupt the last record +my $offset = hex($lastlsn) % $segsize; +open(my $segf, '+<', $walfile) or die "failed to open $walfile\n"; +seek($segf, $offset, 0); # halfway break the last record +print $segf "\0\0\0\0"; +close($segf); + +# pg_waldump complains about the corrupted record +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stderr =~ /fatal: error in WAL record at 0\/$lastlsn: .* at 0\/$lastlsn/, + "pg_waldump emits the correct error message"); + +# also server complains +$logstart = get_log_size($node); +$node->start; +$max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, "WARNING: invalid record length at 0/$lastlsn: wanted [0-9]+, got 0", + $logstart)); + sleep 0.5; +} +ok($max_attempts >= 0, "header error is logged at $lastlsn"); + +# no end-of-wal message should be seen this time +ok(!find_in_log($node, $reached_eow_pat, $logstart), + "false log message is not emitted"); + +$node->stop('immediate'); + done_testing(); + +#### helper routines +# return the size of logfile of $node in bytes +sub get_log_size +{ + my ($node) = @_; + + return (stat $node->logfile)[7]; +} + +# find $pat in logfile of $node after $off-th byte +sub find_in_log +{ + my ($node, $pat, $off) = @_; + + $off = 0 unless defined $off; + my $log = PostgreSQL::Test::Utils::slurp_file($node->logfile); + return 0 if (length($log) <= $off); + + $log = substr($log, $off); + + return $log =~ m/$pat/; +} -- 2.27.0 ----Next_Part(Tue_Mar_29_15_07_01_2022_552)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v5] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlog.c | 94 +++++++++++++++++++------ src/backend/access/transam/xlogreader.c | 14 ++++ src/backend/replication/walreceiver.c | 3 +- src/include/access/xlogreader.h | 1 + 4 files changed, 87 insertions(+), 25 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 5cda30836f..623fb01d0a 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4477,6 +4477,7 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; @@ -4494,6 +4495,16 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * NULL ReadRecPtr means we could not read a record at + * beginning. In that case EndRecPtr is storing the LSN of the + * record we tried to read. + */ + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; } if (readFile >= 0) @@ -4503,13 +4514,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we get here for other than end-of-wal, emit the error message + * right now. Otherwise the message if any is shown as a part of + * the end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4540,11 +4550,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4557,11 +4568,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4609,12 +4626,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * If we haven't emit an error message, we have safely reached the + * end-of-WAL. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(ErrRecPtr), replayTLI, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7544,7 +7582,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false, ThisTimeLineID); + record = ReadRecord(xlogreader, WARNING, false, ThisTimeLineID); } if (record != NULL) @@ -7781,7 +7819,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false, ThisTimeLineID); + record = ReadRecord(xlogreader, WARNING, false, ThisTimeLineID); } while (record != NULL); /* @@ -7841,13 +7879,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -13135,7 +13180,9 @@ XLogShutdownWalRcv(void) * reading from pg_wal, because we don't expect any invalid records in archive * or in records streamed from the primary. Files in the archive should be complete, * and we should never hit the end of WAL because we stop and wait for more WAL - * to arrive before replaying it. + * to arrive before replaying it. When we failed to read a new page, + * readSource is reset to XLOG_FROM_ANY. This indicates all sources including + * pg_wal was failed. Thus treat that the same way with XLOG_FROM_PG_WAL. * * NOTE: This function remembers the RecPtr value it was last called with, * to suppress repeated messages about the same record. Only call this when @@ -13147,7 +13194,8 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if ((readSource == XLOG_FROM_PG_WAL || readSource == XLOG_FROM_ANY) + && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index f39f8044a9..df2198e862 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -121,6 +121,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -292,6 +293,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); state->abortedRecPtr = InvalidXLogRecPtr; @@ -588,6 +590,9 @@ err: */ state->abortedRecPtr = RecPtr; state->missingContrecPtr = targetPagePtr; + report_invalid_record(state, + "missing contrecord at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); } /* @@ -730,6 +735,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index 7a7eb3784e..ba3c4bd550 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -471,8 +471,7 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, LSN_FORMAT_ARGS(LogstreamResult.Write)))); endofwal = true; diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index de6fd791fe..1241b85838 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* * Set at the end of recovery: the start point of a partial record at the -- 2.27.0 ----Next_Part(Mon_Nov__8_14_59_46_2021_800)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v11] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlog.c | 91 +++++++++++++----- src/backend/access/transam/xlogreader.c | 78 +++++++++++++++ src/backend/replication/walreceiver.c | 7 +- src/bin/pg_waldump/pg_waldump.c | 13 ++- src/include/access/xlogreader.h | 1 + src/test/recovery/t/011_crash_recovery.pl | 110 +++++++++++++++++++++- 6 files changed, 270 insertions(+), 30 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index dfe2a0bcce..378c13ccf7 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4480,6 +4480,7 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); if (record == NULL) @@ -4495,6 +4496,18 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * NULL ReadRecPtr means we could not read a record at the + * beginning. In that case EndRecPtr is storing the LSN of the + * record we tried to read. + */ + ErrRecPtr = + xlogreader->ReadRecPtr ? + xlogreader->ReadRecPtr : xlogreader->EndRecPtr; } if (readFile >= 0) @@ -4504,13 +4517,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we get here for other than end-of-wal, emit the error + * message right now. Otherwise the message if any is shown as a + * part of the end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4541,11 +4553,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4558,11 +4571,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4610,12 +4629,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * If we haven't emit an error message, we have safely reached the + * end-of-WAL. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(ErrRecPtr), replayTLI, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7544,7 +7584,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } if (record != NULL) @@ -7782,7 +7822,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } while (record != NULL); /* @@ -7842,13 +7882,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -13097,7 +13144,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 35029cf97d..03a8b42f15 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -121,6 +121,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -292,6 +293,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); state->abortedRecPtr = InvalidXLogRecPtr; @@ -588,6 +590,15 @@ err: */ state->abortedRecPtr = RecPtr; state->missingContrecPtr = targetPagePtr; + + /* + * If the message is not set yet, that means we failed to load the + * page for the record. Otherwise do not hide the existing message. + */ + if (state->errormsg_buf[0] == '\0') + report_invalid_record(state, + "missing contrecord at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); } /* @@ -730,6 +741,40 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* + * We are almost sure reaching the end of WAL, make sure that the + * whole page after the record is filled with zeroes. + */ + char *p; + char *pe; + + /* scan from the beginning of the record to the end of block */ + p = (char *) record; + pe = p + XLOG_BLCKSZ - (RecPtr & (XLOG_BLCKSZ - 1)); + + while (*p == 0 && p < pe) + p++; + + if (p == pe) + { + /* + * The page after the record is completely zeroed. That suggests + * we don't have a record after this point. We don't bother + * checking the pages after since they are not zeroed in the case + * of recycled segments. + */ + report_invalid_record(state, "empty record found at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid record length */ + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, @@ -836,6 +881,31 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, XLogSegNoOffsetToRecPtr(segno, offset, state->segcxt.ws_segsize, recaddr); + StaticAssertStmt(XLOG_PAGE_MAGIC != 0, "XLOG_PAGE_MAGIC is zero"); + + if (hdr->xlp_magic == 0) + { + /* Regard an empty page as End-Of-WAL */ + int i; + + for (i = 0; i < XLOG_BLCKSZ && phdr[i] == 0; i++); + if (i == XLOG_BLCKSZ) + { + char fname[MAXFNAMELEN]; + + XLogFileName(fname, state->seg.ws_tli, segno, + state->segcxt.ws_segsize); + + report_invalid_record(state, + "empty page in log segment %s, offset %u", + fname, + offset); + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid magic number */ + } if (hdr->xlp_magic != XLOG_PAGE_MAGIC) { char fname[MAXFNAMELEN]; @@ -921,6 +991,14 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, LSN_FORMAT_ARGS(hdr->xlp_pageaddr), fname, offset); + + /* + * If the page address is less than expected we assume it is an unused + * page in a recycled segment. + */ + if (hdr->xlp_pageaddr < recaddr) + state->EndOfWAL = true; + return false; } diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index b39fce8c23..8e1fa32489 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -471,10 +471,9 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", - startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", + startpointTLI, + LSN_FORMAT_ARGS(LogstreamResult.Write)))); endofwal = true; break; } diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index a6251e1a96..3745e76488 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -1176,9 +1176,16 @@ main(int argc, char **argv) exit(0); if (errormsg) - fatal_error("error in WAL record at %X/%X: %s", - LSN_FORMAT_ARGS(xlogreader_state->ReadRecPtr), - errormsg); + { + if (xlogreader_state->EndOfWAL) + pg_log_info("end of WAL at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + else + fatal_error("error in WAL record at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + } XLogReaderFree(xlogreader_state); diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 477f0efe26..3eeba220a1 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* * Set at the end of recovery: the start point of a partial record at the diff --git a/src/test/recovery/t/011_crash_recovery.pl b/src/test/recovery/t/011_crash_recovery.pl index 3892aba3e5..67d264df26 100644 --- a/src/test/recovery/t/011_crash_recovery.pl +++ b/src/test/recovery/t/011_crash_recovery.pl @@ -10,9 +10,11 @@ use PostgreSQL::Test::Cluster; use PostgreSQL::Test::Utils; use Test::More; use Config; +use IPC::Run; -plan tests => 3; +plan tests => 11; +my $reached_eow_pat = "reached end of WAL at "; my $node = PostgreSQL::Test::Cluster->new('primary'); $node->init(allows_streaming => 1); $node->start; @@ -50,7 +52,15 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), # Crash and restart the postmaster $node->stop('immediate'); +my $logstart = get_log_size($node); $node->start; +my $max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, $reached_eow_pat, $logstart)); + sleep 0.5; +} +ok ($max_attempts >= 0, "end-of-wal is logged"); # Make sure we really got a new xid cmp_ok($node->safe_psql('postgres', 'SELECT pg_current_xact_id()'), @@ -62,3 +72,101 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), $stdin .= "\\q\n"; $tx->finish; # wait for psql to quit gracefully + +my $segsize = $node->safe_psql('postgres', + qq[SELECT setting FROM pg_settings WHERE name = 'wal_segment_size';]); + +# make sure no records afterwards go to the next segment +$node->safe_psql('postgres', qq[ + SELECT pg_switch_wal(); + CHECKPOINT; + CREATE TABLE t(); +]); +$node->stop('immediate'); + +# identify REDO WAL file +my $cmd = "pg_controldata -D " . $node->data_dir(); +my $chkptfile; +$cmd = ['pg_controldata', '-D', $node->data_dir()]; +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stdout =~ /^Latest checkpoint's REDO WAL file:[ \t] *(.+)$/m, + "checkpoint file is identified"); +my $chkptfile = $1; + +# identify the last record +my $walfile = $node->data_dir() . "/pg_wal/$chkptfile"; +$cmd = ['pg_waldump', $walfile]; +$stdout = ''; +$stderr = ''; +my $lastrec; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +foreach my $l (split(/\r?\n/, $stdout)) +{ + $lastrec = $l; +} +ok(defined $lastrec, "last WAL record is extracted"); +ok($stderr =~ /end of WAL at ([0-9A-F\/]+): .* at \g1/, + "pg_waldump emits the correct ending message"); + +# read the last record LSN excluding leading zeroes +ok ($lastrec =~ /, lsn: 0\/0*([1-9A-F][0-9A-F]+),/, + "LSN of the last record identified"); +my $lastlsn = $1; + +# corrupt the last record +my $offset = hex($lastlsn) % $segsize; +open(my $segf, '+<', $walfile) or die "failed to open $walfile\n"; +seek($segf, $offset, 0); # halfway break the last record +print $segf "\0\0\0\0"; +close($segf); + +# pg_waldump complains about the corrupted record +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stderr =~ /fatal: error in WAL record at 0\/$lastlsn: .* at 0\/$lastlsn/, + "pg_waldump emits the correct error message"); + +# also server complains +$logstart = get_log_size($node); +$node->start; +$max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, "WARNING: invalid record length at 0/$lastlsn: wanted [0-9]+, got 0", + $logstart)); + sleep 0.5; +} +ok($max_attempts >= 0, "header error is logged at $lastlsn"); + +# and the end-of-wal messages shouldn't be seen +# the same message has been confirmed in the past +ok(!find_in_log($node, $reached_eow_pat, $logstart), + "false log message is not emitted"); + +$node->stop('immediate'); + +#### helper routines +# return the size of logfile of $node in bytes +sub get_log_size +{ + my ($node) = @_; + + return (stat $node->logfile)[7]; +} + +# find $pat in logfile of $node after $off-th byte +sub find_in_log +{ + my ($node, $pat, $off) = @_; + + $off = 0 unless defined $off; + my $log = PostgreSQL::Test::Utils::slurp_file($node->logfile); + return 0 if (length($log) <= $off); + + $log = substr($log, $off); + + return $log =~ m/$pat/; +} -- 2.27.0 ----Next_Part(Wed_Feb__2_14_34_58_2022_606)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 45 ++++++++++++++++++++++++++----- 1 file changed, 38 insertions(+), 7 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index d19408b3be..452c376f62 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4288,6 +4288,10 @@ ReadRecord(XLogReaderState *xlogreader, int emode, EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + int actual_emode = + emode_for_corrupt_record(emode, + ReadRecPtr ? ReadRecPtr : EndRecPtr); + if (readFile >= 0) { close(readFile); @@ -4295,14 +4299,41 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * randAccess here means we are reading successive records during + * recovery. If we get here during recovery, we can assume that we + * reached the end of WAL. Otherwise something's really wrong and + * we report just only the errormsg if any. If we don't receive + * errormsg here, we already logged something. We don't emit + * "reached end of WAL" in muted messages. + * + * Note: errormsg is alreay translated. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), - (errmsg_internal("%s", errormsg) /* already translated */ )); + if (!private->randAccess && actual_emode == emode) + { + if (StandbyMode) + ereport(actual_emode, + (errmsg ("rached end of WAL at %X/%X on timeline %u in %s during streaming replication", + (uint32) (EndRecPtr >> 32), (uint32) EndRecPtr, + ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + else if (InArchiveRecovery) + ereport(actual_emode, + (errmsg ("rached end of WAL at %X/%X on timeline %u in %s during archive recovery", + (uint32) (EndRecPtr >> 32), (uint32) EndRecPtr, + ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + else + ereport(actual_emode, + (errmsg ("rached end of WAL at %X/%X on timeline %u in %s during crash recovery", + (uint32) (EndRecPtr >> 32), (uint32) EndRecPtr, + ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } + else if (errormsg) + ereport(actual_emode, (errmsg_internal("%s", errormsg))); } /* -- 2.18.2 ----Next_Part(Fri_Feb_28_16_01_00_2020_273)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v15] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlogreader.c | 125 ++++++++++++++++++---- src/backend/access/transam/xlogrecovery.c | 92 ++++++++++++---- src/backend/replication/walreceiver.c | 7 +- src/bin/pg_waldump/pg_waldump.c | 13 ++- src/include/access/xlogreader.h | 1 + src/test/recovery/t/011_crash_recovery.pl | 106 ++++++++++++++++++ 6 files changed, 297 insertions(+), 47 deletions(-) diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 35029cf97d..ba1c1ece87 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -42,6 +42,8 @@ static bool allocate_recordbuf(XLogReaderState *state, uint32 reclength); static int ReadPageInternal(XLogReaderState *state, XLogRecPtr pageptr, int reqLen); static void XLogReaderInvalReadState(XLogReaderState *state); +static bool ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record); static bool ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess); static bool ValidXLogRecord(XLogReaderState *state, XLogRecord *record, @@ -121,6 +123,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -292,6 +295,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); state->abortedRecPtr = InvalidXLogRecPtr; @@ -380,12 +384,11 @@ restart: * whole header. */ record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); - total_len = record->xl_tot_len; /* * If the whole record header is on this page, validate it immediately. - * Otherwise do just a basic sanity check on xl_tot_len, and validate the - * rest of the header after reading it from the next page. The xl_tot_len + * Otherwise do just a basic sanity check on record length, and validate + * the rest of the header after reading it from the next page. The length * check is necessary here to ensure that we enter the "Need to reassemble * record" code path below; otherwise we might fail to apply * ValidXLogRecordHeader at all. @@ -399,18 +402,13 @@ restart: } else { - /* XXX: more validation should be done here */ - if (total_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: wanted %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, total_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) goto err; - } + gotheader = false; } + total_len = record->xl_tot_len; len = XLOG_BLCKSZ - RecPtr % XLOG_BLCKSZ; if (total_len > len) { @@ -588,6 +586,15 @@ err: */ state->abortedRecPtr = RecPtr; state->missingContrecPtr = targetPagePtr; + + /* + * If the message is not set yet, that means we failed to load the + * page for the record. Otherwise do not hide the existing message. + */ + if (state->errormsg_buf[0] == '\0') + report_invalid_record(state, + "missing contrecord at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); } /* @@ -719,6 +726,60 @@ XLogReaderInvalReadState(XLogReaderState *state) state->readLen = 0; } +/* + * Validate record length of an XLOG record header. + * + * This is substantially a part of ValidXLogRecordHeader. But XLogReadRecord + * needs this separate from the function in case of a partial record header. + */ +static bool +ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record) +{ + if (record->xl_tot_len == 0) + { + char *p; + char *pe; + + /* + * We are almost sure reaching the end of WAL, make sure that the + * whole page after the record is filled with zeroes. + */ + p = (char *) record; + pe = p + XLOG_BLCKSZ - (RecPtr & (XLOG_BLCKSZ - 1)); + + while (*p == 0 && p < pe) + p++; + + if (p == pe) + { + /* + * The page after the record is completely zeroed. That suggests + * we don't have a record after this point. We don't bother + * checking the pages after since they are not zeroed in the case + * of recycled segments. + */ + report_invalid_record(state, "empty record at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + return false; + } + } + + if (record->xl_tot_len < SizeOfXLogRecord) + { + report_invalid_record(state, + "invalid record length at %X/%X: wanted %u, got %u", + LSN_FORMAT_ARGS(RecPtr), + (uint32) SizeOfXLogRecord, record->xl_tot_len); + return false; + } + + return true; +} + /* * Validate an XLOG record header. * @@ -730,14 +791,9 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { - if (record->xl_tot_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: wanted %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, record->xl_tot_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) return false; - } + if (record->xl_rmid > RM_MAX_ID) { report_invalid_record(state, @@ -836,6 +892,31 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, XLogSegNoOffsetToRecPtr(segno, offset, state->segcxt.ws_segsize, recaddr); + StaticAssertStmt(XLOG_PAGE_MAGIC != 0, "XLOG_PAGE_MAGIC is zero"); + + if (hdr->xlp_magic == 0) + { + /* Regard an empty page as End-Of-WAL */ + int i; + + for (i = 0; i < XLOG_BLCKSZ && phdr[i] == 0; i++); + if (i == XLOG_BLCKSZ) + { + char fname[MAXFNAMELEN]; + + XLogFileName(fname, state->seg.ws_tli, segno, + state->segcxt.ws_segsize); + + report_invalid_record(state, + "empty page in log segment %s, offset %u", + fname, + offset); + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid magic number */ + } if (hdr->xlp_magic != XLOG_PAGE_MAGIC) { char fname[MAXFNAMELEN]; @@ -921,6 +1002,14 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, LSN_FORMAT_ARGS(hdr->xlp_pageaddr), fname, offset); + + /* + * If the page address is less than expected we assume it is an unused + * page in a recycled segment. + */ + if (hdr->xlp_pageaddr < recaddr) + state->EndOfWAL = true; + return false; } diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index f9f212680b..750056acaf 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -1592,7 +1592,7 @@ PerformWalRecovery(void) /* just have to read next record after CheckPoint */ Assert(xlogreader->ReadRecPtr == CheckPointLoc); replayTLI = CheckPointTLI; - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } if (record != NULL) @@ -1706,7 +1706,7 @@ PerformWalRecovery(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false, replayTLI); + record = ReadRecord(xlogreader, WARNING, false, replayTLI); } while (record != NULL); /* @@ -1765,13 +1765,20 @@ PerformWalRecovery(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that indicate @@ -2939,6 +2946,7 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); if (record == NULL) @@ -2954,6 +2962,18 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * EndRecPtr is the LSN we tried to read but failed. In the + * case of decoding error, it is at the end of the failed + * record but we don't have a means for now to know EndRecPtr + * is pointing to which of the beginning or ending of the + * failed record. + */ + ErrRecPtr = xlogreader->EndRecPtr; } if (readFile >= 0) @@ -2963,13 +2983,16 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * We only end up here without a message when XLogPageRead() failed + * in that case we already logged something, or just met end-of-WAL + * conditions. In StandbyMode that only happens if we have been + * triggered, so we shouldn't loop anymore in that case. When + * EndOfWAL is true, we don't emit that error if any immediately + * and instead will show it as a part of a decent end-of-wal + * message later. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -3000,11 +3023,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + Assert(ErrRecPtr != InvalidXLogRecPtr); + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -3017,11 +3043,16 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in pg_wal during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -3042,12 +3073,24 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * recovery ended. + * + * Emit a decent message if we met end-of-WAL. Otherwise we should + * have already emitted an error message. + */ + if (xlogreader->EndOfWAL) + ereport(LOG, + (errmsg("reached end of WAL at %X/%X on timeline %u", + LSN_FORMAT_ARGS(ErrRecPtr), replayTLI), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -3129,12 +3172,16 @@ retry: private->replayTLI, xlogreader->EndRecPtr)) { + Assert(!StandbyMode || CheckForStandbyTrigger()); + if (readFile >= 0) close(readFile); readFile = -1; readLen = 0; readSource = XLOG_FROM_ANY; + /* promotion exit is not end-of-WAL */ + xlogreader->EndOfWAL = !StandbyMode; return -1; } } @@ -3767,7 +3814,8 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + /* use currentSource as readSource is reset at failure */ + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index ceaff097b9..4f117ea4da 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -472,10 +472,9 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", - startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); + (errmsg("replication terminated by primary server at %X/%X on timeline %u.", + LSN_FORMAT_ARGS(LogstreamResult.Write), + startpointTLI))); endofwal = true; break; } diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index 2340dc247b..215abe95dc 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -1173,9 +1173,16 @@ main(int argc, char **argv) exit(0); if (errormsg) - fatal_error("error in WAL record at %X/%X: %s", - LSN_FORMAT_ARGS(xlogreader_state->ReadRecPtr), - errormsg); + { + if (xlogreader_state->EndOfWAL) + pg_log_info("end of WAL at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + else + fatal_error("error in WAL record at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + } XLogReaderFree(xlogreader_state); diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 477f0efe26..7b314ef10e 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* was the last attempt EOW? */ /* * Set at the end of recovery: the start point of a partial record at the diff --git a/src/test/recovery/t/011_crash_recovery.pl b/src/test/recovery/t/011_crash_recovery.pl index 14154d1ce0..01033334d6 100644 --- a/src/test/recovery/t/011_crash_recovery.pl +++ b/src/test/recovery/t/011_crash_recovery.pl @@ -10,7 +10,9 @@ use PostgreSQL::Test::Cluster; use PostgreSQL::Test::Utils; use Test::More; use Config; +use IPC::Run; +my $reached_eow_pat = "reached end of WAL at "; my $node = PostgreSQL::Test::Cluster->new('primary'); $node->init(allows_streaming => 1); $node->start; @@ -48,7 +50,15 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), # Crash and restart the postmaster $node->stop('immediate'); +my $logstart = get_log_size($node); $node->start; +my $max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, $reached_eow_pat, $logstart)); + sleep 0.5; +} +ok ($max_attempts >= 0, "end-of-wal is logged"); # Make sure we really got a new xid cmp_ok($node->safe_psql('postgres', 'SELECT pg_current_xact_id()'), @@ -61,4 +71,100 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), $stdin .= "\\q\n"; $tx->finish; # wait for psql to quit gracefully +my $segsize = $node->safe_psql('postgres', + qq[SELECT setting FROM pg_settings WHERE name = 'wal_segment_size';]); + +# make sure no records afterwards go to the next segment +$node->safe_psql('postgres', qq[ + SELECT pg_switch_wal(); + CHECKPOINT; + CREATE TABLE t(); +]); +$node->stop('immediate'); + +# identify REDO WAL file +my $cmd = "pg_controldata -D " . $node->data_dir(); +$cmd = ['pg_controldata', '-D', $node->data_dir()]; +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stdout =~ /^Latest checkpoint's REDO WAL file:[ \t] *(.+)$/m, + "checkpoint file is identified"); +my $chkptfile = $1; + +# identify the last record +my $walfile = $node->data_dir() . "/pg_wal/$chkptfile"; +$cmd = ['pg_waldump', $walfile]; +$stdout = ''; +$stderr = ''; +my $lastrec; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +foreach my $l (split(/\r?\n/, $stdout)) +{ + $lastrec = $l; +} +ok(defined $lastrec, "last WAL record is extracted"); +ok($stderr =~ /end of WAL at ([0-9A-F\/]+): .* at \g1/, + "pg_waldump emits the correct ending message"); + +# read the last record LSN excluding leading zeroes +ok ($lastrec =~ /, lsn: 0\/0*([1-9A-F][0-9A-F]+),/, + "LSN of the last record identified"); +my $lastlsn = $1; + +# corrupt the last record +my $offset = hex($lastlsn) % $segsize; +open(my $segf, '+<', $walfile) or die "failed to open $walfile\n"; +seek($segf, $offset, 0); # halfway break the last record +print $segf "\0\0\0\0"; +close($segf); + +# pg_waldump complains about the corrupted record +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stderr =~ /fatal: error in WAL record at 0\/$lastlsn: .* at 0\/$lastlsn/, + "pg_waldump emits the correct error message"); + +# also server complains +$logstart = get_log_size($node); +$node->start; +$max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, "WARNING: invalid record length at 0/$lastlsn: wanted [0-9]+, got 0", + $logstart)); + sleep 0.5; +} +ok($max_attempts >= 0, "header error is logged at $lastlsn"); + +# no end-of-wal message should be seen this time +ok(!find_in_log($node, $reached_eow_pat, $logstart), + "false log message is not emitted"); + +$node->stop('immediate'); + done_testing(); + +#### helper routines +# return the size of logfile of $node in bytes +sub get_log_size +{ + my ($node) = @_; + + return (stat $node->logfile)[7]; +} + +# find $pat in logfile of $node after $off-th byte +sub find_in_log +{ + my ($node, $pat, $off) = @_; + + $off = 0 unless defined $off; + my $log = PostgreSQL::Test::Utils::slurp_file($node->logfile); + return 0 if (length($log) <= $off); + + $log = substr($log, $off); + + return $log =~ m/$pat/; +} -- 2.27.0 ----Next_Part(Wed_Mar__2_11_17_04_2022_065)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v4] Make End-Of-Recovery error less scary @ 2020-02-28 06:52 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2020-02-28 06:52 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Make this message less scary as "reached end of WAL". --- src/backend/access/transam/xlog.c | 81 ++++++++++++++++++------- src/backend/access/transam/xlogreader.c | 11 ++++ src/backend/replication/walreceiver.c | 13 ++-- src/include/access/xlogreader.h | 1 + 4 files changed, 79 insertions(+), 27 deletions(-) diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 377afb8732..fbcb8d78b8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -4361,12 +4361,15 @@ ReadRecord(XLogReaderState *xlogreader, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogReadRecord(xlogreader, &errormsg); ReadRecPtr = xlogreader->ReadRecPtr; EndRecPtr = xlogreader->EndRecPtr; if (record == NULL) { + ErrRecPtr = ReadRecPtr ? ReadRecPtr : EndRecPtr; + if (readFile >= 0) { close(readFile); @@ -4374,13 +4377,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, } /* - * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * If we met other than end-of-wal, emit the error message right + * now. Otherwise the message if any is shown as a part of the + * end-of-WAL message below. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, EndRecPtr), + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), (errmsg_internal("%s", errormsg) /* already translated */ )); } @@ -4411,11 +4413,12 @@ ReadRecord(XLogReaderState *xlogreader, int emode, /* Great, got a record */ return record; } - else + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -4428,11 +4431,17 @@ ReadRecord(XLogReaderState *xlogreader, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + (errmsg_internal("reached end of WAL at %X/%X on timeline %u in %s during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + ThisTimeLineID, + xlogSourceNames[currentSource]))); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -4480,12 +4489,33 @@ ReadRecord(XLogReaderState *xlogreader, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * We reached the end of WAL, show the messages just once at the + * same LSN. + */ + if (emode_for_corrupt_record(LOG, ErrRecPtr) == LOG) + { + char *fmt; + + if (StandbyMode) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during standby mode"); + else if (InArchiveRecovery) + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during archive recovery"); + else + fmt = gettext_noop("reached end of WAL at %X/%X on timeline %u in %s during crash recovery"); + + ereport(LOG, + (errmsg(fmt, LSN_FORMAT_ARGS(EndRecPtr), ThisTimeLineID, + xlogSourceNames[currentSource]), + (errormsg ? errdetail_internal("%s", errormsg) : 0))); + } } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -7227,7 +7257,7 @@ StartupXLOG(void) else { /* just have to read next record after CheckPoint */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } if (record != NULL) @@ -7454,7 +7484,7 @@ StartupXLOG(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogreader, LOG, false); + record = ReadRecord(xlogreader, WARNING, false); } while (record != NULL); /* @@ -7514,13 +7544,20 @@ StartupXLOG(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, (errmsg("redo is not required"))); } + else + { + /* broken record found */ + ereport(WARNING, + (errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database."))); + } /* * This check is intentionally after the above log messages that @@ -12653,7 +12690,7 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + if (readSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 42738eb940..dacba32143 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -118,6 +118,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -288,6 +289,7 @@ XLogReadRecord(XLogReaderState *state, char **errormsg) /* reset error state */ *errormsg = NULL; state->errormsg_buf[0] = '\0'; + state->EndOfWAL = false; ResetDecoder(state); @@ -689,6 +691,15 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { + if (record->xl_tot_len == 0) + { + /* This is strictly not an invalid state, so phrase it as so. */ + report_invalid_record(state, + "record length is 0 at %X/%X", + (uint32) (RecPtr >> 32), (uint32) RecPtr); + state->EndOfWAL = true; + return false; + } if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index e5f8a06fea..2377c58b4c 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -460,12 +460,15 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", + (errmsg("replication terminated by primary server on timeline %u at %X/%X.", startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); - endofwal = true; - break; + LSN_FORMAT_ARGS(LogstreamResult.Write)))); + + /* + * we have no longer anything to do on the broken + * connection other than exiting. + */ + proc_exit(1); } len = walrcv_receive(wrconn, &buf, &wait_fd); } diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 21d200d3df..0491adfc5b 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -174,6 +174,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* the last attempt was EOW? */ /* ---------------------------------------- -- 2.27.0 ----Next_Part(Thu_Mar__4_15_50_39_2021_269)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v21] Make End-Of-Recovery error less scary @ 2022-07-07 02:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2022-07-07 02:51 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlogreader.c | 136 +++++++++++++++++----- src/backend/access/transam/xlogrecovery.c | 94 +++++++++++---- src/backend/replication/walreceiver.c | 7 +- src/bin/pg_waldump/pg_waldump.c | 13 ++- src/include/access/xlogreader.h | 1 + src/test/recovery/t/011_crash_recovery.pl | 106 +++++++++++++++++ 6 files changed, 298 insertions(+), 59 deletions(-) diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 4d6c34e0fc..b03eeb1487 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -48,6 +48,8 @@ static int ReadPageInternal(XLogReaderState *state, XLogRecPtr pageptr, int reqLen); static void XLogReaderInvalReadState(XLogReaderState *state); static XLogPageReadResult XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking); +static bool ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record); static bool ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess); static bool ValidXLogRecord(XLogReaderState *state, XLogRecord *record, @@ -149,6 +151,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -558,6 +561,7 @@ XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking) /* reset error state */ state->errormsg_buf[0] = '\0'; decoded = NULL; + state->EndOfWAL = false; state->abortedRecPtr = InvalidXLogRecPtr; state->missingContrecPtr = InvalidXLogRecPtr; @@ -640,25 +644,21 @@ restart: Assert(pageHeaderSize <= readOff); /* - * Read the record length. + * Validate the record header. * - * NB: Even though we use an XLogRecord pointer here, the whole record - * header might not fit on this page. xl_tot_len is the first field of the - * struct, so it must be on this page (the records are MAXALIGNed), but we - * cannot access any other fields until we've verified that we got the - * whole header. - */ - record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); - total_len = record->xl_tot_len; - - /* - * If the whole record header is on this page, validate it immediately. - * Otherwise do just a basic sanity check on xl_tot_len, and validate the - * rest of the header after reading it from the next page. The xl_tot_len + * Even though we use an XLogRecord pointer here, the whole record header + * might not fit on this page. If the whole record header is on this page, + * validate it immediately. Even otherwise xl_tot_len must be on this page + * (it is the first field of MAXALIGNed records), but we still cannot + * access any further fields until we've verified that we got the whole + * header, so do just a basic sanity check on record length, and validate + * the rest of the header after reading it from the next page. The length * check is necessary here to ensure that we enter the "Need to reassemble * record" code path below; otherwise we might fail to apply * ValidXLogRecordHeader at all. */ + record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); + if (targetRecOff <= XLOG_BLCKSZ - SizeOfXLogRecord) { if (!ValidXLogRecordHeader(state, RecPtr, state->DecodeRecPtr, record, @@ -668,18 +668,14 @@ restart: } else { - /* XXX: more validation should be done here */ - if (total_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: wanted %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, total_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) goto err; - } + gotheader = false; } + total_len = record->xl_tot_len; + /* * Find space to decode this record. Don't allow oversized allocation if * the caller requested nonblocking. Otherwise, we *have* to try to @@ -1105,6 +1101,60 @@ XLogReaderInvalReadState(XLogReaderState *state) state->readLen = 0; } +/* + * Validate record length of an XLOG record header. + * + * This is substantially a part of ValidXLogRecordHeader. But XLogReadRecord + * needs this separate from the function in case of a partial record header. + */ +static bool +ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record) +{ + if (record->xl_tot_len == 0) + { + char *p; + char *pe; + + /* + * We are almost sure reaching the end of WAL, make sure that the + * whole page after the record is filled with zeroes. + */ + p = (char *) record; + pe = p + XLOG_BLCKSZ - (RecPtr & (XLOG_BLCKSZ - 1)); + + while (*p == 0 && p < pe) + p++; + + if (p == pe) + { + /* + * The page after the record is completely zeroed. That suggests + * we don't have a record after this point. We don't bother + * checking the pages after since they are not zeroed in the case + * of recycled segments. + */ + report_invalid_record(state, "empty record at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + return false; + } + } + + if (record->xl_tot_len < SizeOfXLogRecord) + { + report_invalid_record(state, + "invalid record length at %X/%X: wanted %u, got %u", + LSN_FORMAT_ARGS(RecPtr), + (uint32) SizeOfXLogRecord, record->xl_tot_len); + return false; + } + + return true; +} + /* * Validate an XLOG record header. * @@ -1116,14 +1166,9 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { - if (record->xl_tot_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: wanted %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, record->xl_tot_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) return false; - } + if (!RmgrIdIsValid(record->xl_rmid)) { report_invalid_record(state, @@ -1222,6 +1267,31 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, XLogSegNoOffsetToRecPtr(segno, offset, state->segcxt.ws_segsize, recaddr); + StaticAssertStmt(XLOG_PAGE_MAGIC != 0, "XLOG_PAGE_MAGIC is zero"); + + if (hdr->xlp_magic == 0) + { + /* Regard an empty page as End-Of-WAL */ + int i; + + for (i = 0; i < XLOG_BLCKSZ && phdr[i] == 0; i++); + if (i == XLOG_BLCKSZ) + { + char fname[MAXFNAMELEN]; + + XLogFileName(fname, state->seg.ws_tli, segno, + state->segcxt.ws_segsize); + + report_invalid_record(state, + "empty page in log segment %s, offset %u", + fname, + offset); + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid magic number */ + } if (hdr->xlp_magic != XLOG_PAGE_MAGIC) { char fname[MAXFNAMELEN]; @@ -1307,6 +1377,14 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, LSN_FORMAT_ARGS(hdr->xlp_pageaddr), fname, offset); + + /* + * If the page address is less than expected we assume it is an unused + * page in a recycled segment. + */ + if (hdr->xlp_pageaddr < recaddr) + state->EndOfWAL = true; + return false; } diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index b41e682664..56f29e73fe 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -1626,7 +1626,7 @@ PerformWalRecovery(void) /* just have to read next record after CheckPoint */ Assert(xlogreader->ReadRecPtr == CheckPointLoc); replayTLI = CheckPointTLI; - record = ReadRecord(xlogprefetcher, LOG, false, replayTLI); + record = ReadRecord(xlogprefetcher, WARNING, false, replayTLI); } if (record != NULL) @@ -1735,7 +1735,7 @@ PerformWalRecovery(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogprefetcher, LOG, false, replayTLI); + record = ReadRecord(xlogprefetcher, WARNING, false, replayTLI); } while (record != NULL); /* @@ -1789,11 +1789,18 @@ PerformWalRecovery(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, - (errmsg("redo is not required"))); + errmsg("redo is not required")); + } + else + { + /* broken record found */ + ereport(WARNING, + errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database.")); } /* @@ -3024,6 +3031,7 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogPrefetcherReadRecord(xlogprefetcher, &errormsg); if (record == NULL) @@ -3045,6 +3053,18 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * EndRecPtr is the LSN we tried to read but failed. In the + * case of decoding error, it is at the end of the failed + * record but we don't have a means for now to know EndRecPtr + * is pointing to which of the beginning or ending of the + * failed record. + */ + ErrRecPtr = xlogreader->EndRecPtr; } if (readFile >= 0) @@ -3055,13 +3075,15 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, /* * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * failed- in that case we already logged something. In StandbyMode + * that only happens if we have been triggered, so we shouldn't + * loop anymore in that case. When EndOfWAL is true, we don't emit + * the message immediately and instead will show it as a part of a + * decent end-of-wal message later. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), - (errmsg_internal("%s", errormsg) /* already translated */ )); + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), + errmsg_internal("%s", errormsg) /* already translated */ ); } /* @@ -3091,11 +3113,14 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, /* Great, got a record */ return record; } - else + + Assert(ErrRecPtr != InvalidXLogRecPtr); + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -3108,11 +3133,16 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + errmsg_internal("reached end of WAL at %X/%X on timeline %u in pg_wal during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI)); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -3133,12 +3163,24 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * recovery ended. + * + * Emit a decent message if we met end-of-WAL. Otherwise we should + * have already emitted an error message. + */ + if (xlogreader->EndOfWAL) + ereport(LOG, + errmsg("reached end of WAL at %X/%X on timeline %u", + LSN_FORMAT_ARGS(ErrRecPtr), replayTLI), + (errormsg ? errdetail_internal("%s", errormsg) : 0)); } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -3233,11 +3275,16 @@ retry: case XLREAD_WOULDBLOCK: return XLREAD_WOULDBLOCK; case XLREAD_FAIL: + Assert(!StandbyMode || CheckForStandbyTrigger()); + if (readFile >= 0) close(readFile); readFile = -1; readLen = 0; readSource = XLOG_FROM_ANY; + + /* promotion exit is not end-of-WAL */ + xlogreader->EndOfWAL = !StandbyMode; return XLREAD_FAIL; case XLREAD_SUCCESS: break; @@ -3898,7 +3945,8 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + /* use currentSource as readSource is reset at failure */ + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index f6ef0ace2c..c6d7be6688 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -472,10 +472,9 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", - startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); + errmsg("replication terminated by primary server at %X/%X on timeline %u.", + LSN_FORMAT_ARGS(LogstreamResult.Write), + startpointTLI)); endofwal = true; break; } diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index 9993378ca5..26a4125b30 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -1168,9 +1168,16 @@ main(int argc, char **argv) exit(0); if (errormsg) - pg_fatal("error in WAL record at %X/%X: %s", - LSN_FORMAT_ARGS(xlogreader_state->ReadRecPtr), - errormsg); + { + if (xlogreader_state->EndOfWAL) + pg_log_info("end of WAL at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + else + pg_fatal("error in WAL record at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + } XLogReaderFree(xlogreader_state); diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 6dcde2523a..0818cb7ef0 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -205,6 +205,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* was the last attempt EOW? */ /* * Set at the end of recovery: the start point of a partial record at the diff --git a/src/test/recovery/t/011_crash_recovery.pl b/src/test/recovery/t/011_crash_recovery.pl index 1b57d01046..bde16b7cfa 100644 --- a/src/test/recovery/t/011_crash_recovery.pl +++ b/src/test/recovery/t/011_crash_recovery.pl @@ -9,7 +9,9 @@ use warnings; use PostgreSQL::Test::Cluster; use PostgreSQL::Test::Utils; use Test::More; +use IPC::Run; +my $reached_eow_pat = "reached end of WAL at "; my $node = PostgreSQL::Test::Cluster->new('primary'); $node->init(allows_streaming => 1); $node->start; @@ -47,7 +49,15 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), # Crash and restart the postmaster $node->stop('immediate'); +my $logstart = get_log_size($node); $node->start; +my $max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, $reached_eow_pat, $logstart)); + sleep 0.5; +} +ok ($max_attempts >= 0, "end-of-wal is logged"); # Make sure we really got a new xid cmp_ok($node->safe_psql('postgres', 'SELECT pg_current_xact_id()'), @@ -60,4 +70,100 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), $stdin .= "\\q\n"; $tx->finish; # wait for psql to quit gracefully +my $segsize = $node->safe_psql('postgres', + qq[SELECT setting FROM pg_settings WHERE name = 'wal_segment_size';]); + +# make sure no records afterwards go to the next segment +$node->safe_psql('postgres', qq[ + SELECT pg_switch_wal(); + CHECKPOINT; + CREATE TABLE t(); +]); +$node->stop('immediate'); + +# identify REDO WAL file +my $cmd = "pg_controldata -D " . $node->data_dir(); +$cmd = ['pg_controldata', '-D', $node->data_dir()]; +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stdout =~ /^Latest checkpoint's REDO WAL file:[ \t] *(.+)$/m, + "checkpoint file is identified"); +my $chkptfile = $1; + +# identify the last record +my $walfile = $node->data_dir() . "/pg_wal/$chkptfile"; +$cmd = ['pg_waldump', $walfile]; +$stdout = ''; +$stderr = ''; +my $lastrec; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +foreach my $l (split(/\r?\n/, $stdout)) +{ + $lastrec = $l; +} +ok(defined $lastrec, "last WAL record is extracted"); +ok($stderr =~ /end of WAL at ([0-9A-F\/]+): .* at \g1/, + "pg_waldump emits the correct ending message"); + +# read the last record LSN excluding leading zeroes +ok ($lastrec =~ /, lsn: 0\/0*([1-9A-F][0-9A-F]+),/, + "LSN of the last record identified"); +my $lastlsn = $1; + +# corrupt the last record +my $offset = hex($lastlsn) % $segsize; +open(my $segf, '+<', $walfile) or die "failed to open $walfile\n"; +seek($segf, $offset, 0); # halfway corrupt the last record +print $segf "\0\0\0\0"; +close($segf); + +# pg_waldump complains about the corrupted record +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stderr =~ /error: error in WAL record at 0\/$lastlsn: .* at 0\/$lastlsn/, + "pg_waldump emits the correct error message"); + +# also server complains +$logstart = get_log_size($node); +$node->start; +$max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, "WARNING: invalid record length at 0/$lastlsn: wanted [0-9]+, got 0", + $logstart)); + sleep 0.5; +} +ok($max_attempts >= 0, "header error is logged at $lastlsn"); + +# no end-of-wal message should be seen this time +ok(!find_in_log($node, $reached_eow_pat, $logstart), + "false log message is not emitted"); + +$node->stop('immediate'); + done_testing(); + +#### helper routines +# return the size of logfile of $node in bytes +sub get_log_size +{ + my ($node) = @_; + + return (stat $node->logfile)[7]; +} + +# find $pat in logfile of $node after $off-th byte +sub find_in_log +{ + my ($node, $pat, $off) = @_; + + $off = 0 unless defined $off; + my $log = PostgreSQL::Test::Utils::slurp_file($node->logfile); + return 0 if (length($log) <= $off); + + $log = substr($log, $off); + + return $log =~ m/$pat/; +} -- 2.31.1 ----Next_Part(Mon_Sep_26_16_17_37_2022_881)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v19] Make End-Of-Recovery error less scary @ 2022-07-07 02:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2022-07-07 02:51 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlogreader.c | 145 +++++++++++++++++----- src/backend/access/transam/xlogrecovery.c | 95 ++++++++++---- src/backend/replication/walreceiver.c | 7 +- src/bin/pg_waldump/pg_waldump.c | 13 +- src/include/access/xlogreader.h | 1 + src/test/recovery/t/011_crash_recovery.pl | 106 ++++++++++++++++ 6 files changed, 308 insertions(+), 59 deletions(-) diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index f3dc4b7797..a86ab2b02b 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -48,6 +48,8 @@ static int ReadPageInternal(XLogReaderState *state, XLogRecPtr pageptr, int reqLen); static void XLogReaderInvalReadState(XLogReaderState *state); static XLogPageReadResult XLogDecodeNextRecord(XLogReaderState *state, bool non_blocking); +static bool ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record); static bool ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess); static bool ValidXLogRecord(XLogReaderState *state, XLogRecord *record, @@ -149,6 +151,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -554,6 +557,7 @@ XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking) /* reset error state */ state->errormsg_buf[0] = '\0'; decoded = NULL; + state->EndOfWAL = false; state->abortedRecPtr = InvalidXLogRecPtr; state->missingContrecPtr = InvalidXLogRecPtr; @@ -635,25 +639,21 @@ restart: Assert(pageHeaderSize <= readOff); /* - * Read the record length. + * Validate the record header. * - * NB: Even though we use an XLogRecord pointer here, the whole record - * header might not fit on this page. xl_tot_len is the first field of the - * struct, so it must be on this page (the records are MAXALIGNed), but we - * cannot access any other fields until we've verified that we got the - * whole header. - */ - record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); - total_len = record->xl_tot_len; - - /* - * If the whole record header is on this page, validate it immediately. - * Otherwise do just a basic sanity check on xl_tot_len, and validate the - * rest of the header after reading it from the next page. The xl_tot_len + * Even though we use an XLogRecord pointer here, the whole record header + * might not fit on this page. If the whole record header is on this page, + * validate it immediately. Even otherwise xl_tot_len must be on this page + * (it is the first field of MAXALIGNed records), but we still cannot + * access any further fields until we've verified that we got the whole + * header, so do just a basic sanity check on record length, and validate + * the rest of the header after reading it from the next page. The length * check is necessary here to ensure that we enter the "Need to reassemble * record" code path below; otherwise we might fail to apply * ValidXLogRecordHeader at all. */ + record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); + if (targetRecOff <= XLOG_BLCKSZ - SizeOfXLogRecord) { if (!ValidXLogRecordHeader(state, RecPtr, state->DecodeRecPtr, record, @@ -663,18 +663,14 @@ restart: } else { - /* XXX: more validation should be done here */ - if (total_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: wanted %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, total_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) goto err; - } + gotheader = false; } + total_len = record->xl_tot_len; + /* * Find space to decode this record. Don't allow oversized allocation if * the caller requested nonblocking. Otherwise, we *have* to try to @@ -906,6 +902,15 @@ err: */ state->abortedRecPtr = RecPtr; state->missingContrecPtr = targetPagePtr; + + /* + * If the message is not set yet, that means we failed to load the + * page for the record. Otherwise do not hide the existing message. + */ + if (state->errormsg_buf[0] == '\0') + report_invalid_record(state, + "missing contrecord at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); } if (decoded && decoded->oversized) @@ -1085,6 +1090,60 @@ XLogReaderInvalReadState(XLogReaderState *state) state->readLen = 0; } +/* + * Validate record length of an XLOG record header. + * + * This is substantially a part of ValidXLogRecordHeader. But XLogReadRecord + * needs this separate from the function in case of a partial record header. + */ +static bool +ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record) +{ + if (record->xl_tot_len == 0) + { + char *p; + char *pe; + + /* + * We are almost sure reaching the end of WAL, make sure that the + * whole page after the record is filled with zeroes. + */ + p = (char *) record; + pe = p + XLOG_BLCKSZ - (RecPtr & (XLOG_BLCKSZ - 1)); + + while (*p == 0 && p < pe) + p++; + + if (p == pe) + { + /* + * The page after the record is completely zeroed. That suggests + * we don't have a record after this point. We don't bother + * checking the pages after since they are not zeroed in the case + * of recycled segments. + */ + report_invalid_record(state, "empty record at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + return false; + } + } + + if (record->xl_tot_len < SizeOfXLogRecord) + { + report_invalid_record(state, + "invalid record length at %X/%X: wanted %u, got %u", + LSN_FORMAT_ARGS(RecPtr), + (uint32) SizeOfXLogRecord, record->xl_tot_len); + return false; + } + + return true; +} + /* * Validate an XLOG record header. * @@ -1096,14 +1155,9 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { - if (record->xl_tot_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: wanted %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, record->xl_tot_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) return false; - } + if (!RmgrIdIsValid(record->xl_rmid)) { report_invalid_record(state, @@ -1202,6 +1256,31 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, XLogSegNoOffsetToRecPtr(segno, offset, state->segcxt.ws_segsize, recaddr); + StaticAssertStmt(XLOG_PAGE_MAGIC != 0, "XLOG_PAGE_MAGIC is zero"); + + if (hdr->xlp_magic == 0) + { + /* Regard an empty page as End-Of-WAL */ + int i; + + for (i = 0; i < XLOG_BLCKSZ && phdr[i] == 0; i++); + if (i == XLOG_BLCKSZ) + { + char fname[MAXFNAMELEN]; + + XLogFileName(fname, state->seg.ws_tli, segno, + state->segcxt.ws_segsize); + + report_invalid_record(state, + "empty page in log segment %s, offset %u", + fname, + offset); + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid magic number */ + } if (hdr->xlp_magic != XLOG_PAGE_MAGIC) { char fname[MAXFNAMELEN]; @@ -1287,6 +1366,14 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, LSN_FORMAT_ARGS(hdr->xlp_pageaddr), fname, offset); + + /* + * If the page address is less than expected we assume it is an unused + * page in a recycled segment. + */ + if (hdr->xlp_pageaddr < recaddr) + state->EndOfWAL = true; + return false; } diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index 5d6f1b5e46..baae4e84cf 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -1622,7 +1622,7 @@ PerformWalRecovery(void) /* just have to read next record after CheckPoint */ Assert(xlogreader->ReadRecPtr == CheckPointLoc); replayTLI = CheckPointTLI; - record = ReadRecord(xlogprefetcher, LOG, false, replayTLI); + record = ReadRecord(xlogprefetcher, WARNING, false, replayTLI); } if (record != NULL) @@ -1731,7 +1731,7 @@ PerformWalRecovery(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogprefetcher, LOG, false, replayTLI); + record = ReadRecord(xlogprefetcher, WARNING, false, replayTLI); } while (record != NULL); /* @@ -1785,11 +1785,19 @@ PerformWalRecovery(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, - (errmsg("redo is not required"))); + errmsg("redo is not required")); + + } + else + { + /* broken record found */ + ereport(WARNING, + errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database.")); } /* @@ -2969,6 +2977,7 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogPrefetcherReadRecord(xlogprefetcher, &errormsg); if (record == NULL) @@ -2984,6 +2993,18 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * EndRecPtr is the LSN we tried to read but failed. In the + * case of decoding error, it is at the end of the failed + * record but we don't have a means for now to know EndRecPtr + * is pointing to which of the beginning or ending of the + * failed record. + */ + ErrRecPtr = xlogreader->EndRecPtr; } if (readFile >= 0) @@ -2994,13 +3015,15 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, /* * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * failed- in that case we already logged something. In StandbyMode + * that only happens if we have been triggered, so we shouldn't + * loop anymore in that case. When EndOfWAL is true, we don't emit + * the message immediately and instead will show it as a part of a + * decent end-of-wal message later. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), - (errmsg_internal("%s", errormsg) /* already translated */ )); + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), + errmsg_internal("%s", errormsg) /* already translated */ ); } /* @@ -3030,11 +3053,14 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, /* Great, got a record */ return record; } - else + + Assert(ErrRecPtr != InvalidXLogRecPtr); + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -3047,11 +3073,16 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + errmsg_internal("reached end of WAL at %X/%X on timeline %u in pg_wal during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI)); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -3072,12 +3103,24 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * recovery ended. + * + * Emit a decent message if we met end-of-WAL. Otherwise we should + * have already emitted an error message. + */ + if (xlogreader->EndOfWAL) + ereport(LOG, + errmsg("reached end of WAL at %X/%X on timeline %u", + LSN_FORMAT_ARGS(ErrRecPtr), replayTLI), + (errormsg ? errdetail_internal("%s", errormsg) : 0)); } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -3172,11 +3215,16 @@ retry: case XLREAD_WOULDBLOCK: return XLREAD_WOULDBLOCK; case XLREAD_FAIL: + Assert(!StandbyMode || CheckForStandbyTrigger()); + if (readFile >= 0) close(readFile); readFile = -1; readLen = 0; readSource = XLOG_FROM_ANY; + + /* promotion exit is not end-of-WAL */ + xlogreader->EndOfWAL = !StandbyMode; return XLREAD_FAIL; case XLREAD_SUCCESS: break; @@ -3830,7 +3878,8 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + /* use currentSource as readSource is reset at failure */ + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index 3d37c1fe62..73f7641c4f 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -472,10 +472,9 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", - startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); + errmsg("replication terminated by primary server at %X/%X on timeline %u.", + LSN_FORMAT_ARGS(LogstreamResult.Write), + startpointTLI)); endofwal = true; break; } diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index 6528113628..bd09a62a9d 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -1168,9 +1168,16 @@ main(int argc, char **argv) exit(0); if (errormsg) - pg_fatal("error in WAL record at %X/%X: %s", - LSN_FORMAT_ARGS(xlogreader_state->ReadRecPtr), - errormsg); + { + if (xlogreader_state->EndOfWAL) + pg_log_info("end of WAL at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + else + pg_fatal("error in WAL record at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + } XLogReaderFree(xlogreader_state); diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 5395f155aa..b2dae53557 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -205,6 +205,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* was the last attempt EOW? */ /* * Set at the end of recovery: the start point of a partial record at the diff --git a/src/test/recovery/t/011_crash_recovery.pl b/src/test/recovery/t/011_crash_recovery.pl index 1b57d01046..bde16b7cfa 100644 --- a/src/test/recovery/t/011_crash_recovery.pl +++ b/src/test/recovery/t/011_crash_recovery.pl @@ -9,7 +9,9 @@ use warnings; use PostgreSQL::Test::Cluster; use PostgreSQL::Test::Utils; use Test::More; +use IPC::Run; +my $reached_eow_pat = "reached end of WAL at "; my $node = PostgreSQL::Test::Cluster->new('primary'); $node->init(allows_streaming => 1); $node->start; @@ -47,7 +49,15 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), # Crash and restart the postmaster $node->stop('immediate'); +my $logstart = get_log_size($node); $node->start; +my $max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, $reached_eow_pat, $logstart)); + sleep 0.5; +} +ok ($max_attempts >= 0, "end-of-wal is logged"); # Make sure we really got a new xid cmp_ok($node->safe_psql('postgres', 'SELECT pg_current_xact_id()'), @@ -60,4 +70,100 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), $stdin .= "\\q\n"; $tx->finish; # wait for psql to quit gracefully +my $segsize = $node->safe_psql('postgres', + qq[SELECT setting FROM pg_settings WHERE name = 'wal_segment_size';]); + +# make sure no records afterwards go to the next segment +$node->safe_psql('postgres', qq[ + SELECT pg_switch_wal(); + CHECKPOINT; + CREATE TABLE t(); +]); +$node->stop('immediate'); + +# identify REDO WAL file +my $cmd = "pg_controldata -D " . $node->data_dir(); +$cmd = ['pg_controldata', '-D', $node->data_dir()]; +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stdout =~ /^Latest checkpoint's REDO WAL file:[ \t] *(.+)$/m, + "checkpoint file is identified"); +my $chkptfile = $1; + +# identify the last record +my $walfile = $node->data_dir() . "/pg_wal/$chkptfile"; +$cmd = ['pg_waldump', $walfile]; +$stdout = ''; +$stderr = ''; +my $lastrec; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +foreach my $l (split(/\r?\n/, $stdout)) +{ + $lastrec = $l; +} +ok(defined $lastrec, "last WAL record is extracted"); +ok($stderr =~ /end of WAL at ([0-9A-F\/]+): .* at \g1/, + "pg_waldump emits the correct ending message"); + +# read the last record LSN excluding leading zeroes +ok ($lastrec =~ /, lsn: 0\/0*([1-9A-F][0-9A-F]+),/, + "LSN of the last record identified"); +my $lastlsn = $1; + +# corrupt the last record +my $offset = hex($lastlsn) % $segsize; +open(my $segf, '+<', $walfile) or die "failed to open $walfile\n"; +seek($segf, $offset, 0); # halfway corrupt the last record +print $segf "\0\0\0\0"; +close($segf); + +# pg_waldump complains about the corrupted record +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stderr =~ /error: error in WAL record at 0\/$lastlsn: .* at 0\/$lastlsn/, + "pg_waldump emits the correct error message"); + +# also server complains +$logstart = get_log_size($node); +$node->start; +$max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, "WARNING: invalid record length at 0/$lastlsn: wanted [0-9]+, got 0", + $logstart)); + sleep 0.5; +} +ok($max_attempts >= 0, "header error is logged at $lastlsn"); + +# no end-of-wal message should be seen this time +ok(!find_in_log($node, $reached_eow_pat, $logstart), + "false log message is not emitted"); + +$node->stop('immediate'); + done_testing(); + +#### helper routines +# return the size of logfile of $node in bytes +sub get_log_size +{ + my ($node) = @_; + + return (stat $node->logfile)[7]; +} + +# find $pat in logfile of $node after $off-th byte +sub find_in_log +{ + my ($node, $pat, $off) = @_; + + $off = 0 unless defined $off; + my $log = PostgreSQL::Test::Utils::slurp_file($node->logfile); + return 0 if (length($log) <= $off); + + $log = substr($log, $off); + + return $log =~ m/$pat/; +} -- 2.31.1 ----Next_Part(Thu_Jul__7_17_32_33_2022_801)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH] Make End-Of-Recovery error less scary @ 2022-07-07 02:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2022-07-07 02:51 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlogreader.c | 135 +++++++++++++++++----- src/backend/access/transam/xlogrecovery.c | 95 +++++++++++---- src/backend/replication/walreceiver.c | 7 +- src/bin/pg_waldump/pg_waldump.c | 13 ++- src/include/access/xlogreader.h | 1 + src/test/recovery/t/011_crash_recovery.pl | 106 +++++++++++++++++ 6 files changed, 299 insertions(+), 58 deletions(-) diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 93f667b2544..f891a629443 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -48,6 +48,8 @@ static int ReadPageInternal(XLogReaderState *state, XLogRecPtr pageptr, int reqLen); static void XLogReaderInvalReadState(XLogReaderState *state); static XLogPageReadResult XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking); +static bool ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record); static bool ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess); static bool ValidXLogRecord(XLogReaderState *state, XLogRecord *record, @@ -149,6 +151,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -558,6 +561,7 @@ XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking) /* reset error state */ state->errormsg_buf[0] = '\0'; decoded = NULL; + state->EndOfWAL = false; state->abortedRecPtr = InvalidXLogRecPtr; state->missingContrecPtr = InvalidXLogRecPtr; @@ -640,25 +644,21 @@ restart: Assert(pageHeaderSize <= readOff); /* - * Read the record length. + * Validate the record header. * - * NB: Even though we use an XLogRecord pointer here, the whole record - * header might not fit on this page. xl_tot_len is the first field of the - * struct, so it must be on this page (the records are MAXALIGNed), but we - * cannot access any other fields until we've verified that we got the - * whole header. - */ - record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); - total_len = record->xl_tot_len; - - /* - * If the whole record header is on this page, validate it immediately. - * Otherwise do just a basic sanity check on xl_tot_len, and validate the - * rest of the header after reading it from the next page. The xl_tot_len + * Even though we use an XLogRecord pointer here, the whole record header + * might not fit on this page. If the whole record header is on this page, + * validate it immediately. Even otherwise xl_tot_len must be on this page + * (it is the first field of MAXALIGNed records), but we still cannot + * access any further fields until we've verified that we got the whole + * header, so do just a basic sanity check on record length, and validate + * the rest of the header after reading it from the next page. The length * check is necessary here to ensure that we enter the "Need to reassemble * record" code path below; otherwise we might fail to apply * ValidXLogRecordHeader at all. */ + record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); + if (targetRecOff <= XLOG_BLCKSZ - SizeOfXLogRecord) { if (!ValidXLogRecordHeader(state, RecPtr, state->DecodeRecPtr, record, @@ -668,18 +668,14 @@ restart: } else { - /* XXX: more validation should be done here */ - if (total_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: wanted %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, total_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) goto err; - } + gotheader = false; } + total_len = record->xl_tot_len; + /* * Find space to decode this record. Don't allow oversized allocation if * the caller requested nonblocking. Otherwise, we *have* to try to @@ -1106,16 +1102,47 @@ XLogReaderInvalReadState(XLogReaderState *state) } /* - * Validate an XLOG record header. + * Validate record length of an XLOG record header. * - * This is just a convenience subroutine to avoid duplicated code in - * XLogReadRecord. It's not intended for use from anywhere else. + * This is substantially a part of ValidXLogRecordHeader. But XLogReadRecord + * needs this separate from the function in case of a partial record header. */ static bool -ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, - XLogRecPtr PrevRecPtr, XLogRecord *record, - bool randAccess) +ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record) { + if (record->xl_tot_len == 0) + { + char *p; + char *pe; + + /* + * We are almost sure reaching the end of WAL, make sure that the + * whole page after the record is filled with zeroes. + */ + p = (char *) record; + pe = p + XLOG_BLCKSZ - (RecPtr & (XLOG_BLCKSZ - 1)); + + while (*p == 0 && p < pe) + p++; + + if (p == pe) + { + /* + * The page after the record is completely zeroed. That suggests + * we don't have a record after this point. We don't bother + * checking the pages after since they are not zeroed in the case + * of recycled segments. + */ + report_invalid_record(state, "empty record at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + return false; + } + } + if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, @@ -1124,6 +1151,24 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, (uint32) SizeOfXLogRecord, record->xl_tot_len); return false; } + + return true; +} + +/* + * Validate an XLOG record header. + * + * This is just a convenience subroutine to avoid duplicated code in + * XLogReadRecord. It's not intended for use from anywhere else. + */ +static bool +ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecPtr PrevRecPtr, XLogRecord *record, + bool randAccess) +{ + if (!ValidXLogRecordLength(state, RecPtr, record)) + return false; + if (!RmgrIdIsValid(record->xl_rmid)) { report_invalid_record(state, @@ -1219,6 +1264,32 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, XLByteToSeg(recptr, segno, state->segcxt.ws_segsize); offset = XLogSegmentOffset(recptr, state->segcxt.ws_segsize); + StaticAssertStmt(XLOG_PAGE_MAGIC != 0, "XLOG_PAGE_MAGIC is zero"); + + if (hdr->xlp_magic == 0) + { + /* Regard an empty page as End-Of-WAL */ + int i; + + for (i = 0; i < XLOG_BLCKSZ && phdr[i] == 0; i++); + if (i == XLOG_BLCKSZ) + { + char fname[MAXFNAMELEN]; + + XLogFileName(fname, state->seg.ws_tli, segno, + state->segcxt.ws_segsize); + + report_invalid_record(state, + "empty page in log segment %s, offset %u", + fname, + offset); + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid magic number */ + } + if (hdr->xlp_magic != XLOG_PAGE_MAGIC) { char fname[MAXFNAMELEN]; @@ -1304,6 +1375,14 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, LSN_FORMAT_ARGS(hdr->xlp_pageaddr), fname, offset); + + /* + * If the page address is less than expected we assume it is an unused + * page in a recycled segment. + */ + if (hdr->xlp_pageaddr < recptr) + state->EndOfWAL = true; + return false; } diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index cb07694aea6..0034f65af02 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -1626,7 +1626,7 @@ PerformWalRecovery(void) /* just have to read next record after CheckPoint */ Assert(xlogreader->ReadRecPtr == CheckPointLoc); replayTLI = CheckPointTLI; - record = ReadRecord(xlogprefetcher, LOG, false, replayTLI); + record = ReadRecord(xlogprefetcher, WARNING, false, replayTLI); } if (record != NULL) @@ -1735,7 +1735,7 @@ PerformWalRecovery(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogprefetcher, LOG, false, replayTLI); + record = ReadRecord(xlogprefetcher, WARNING, false, replayTLI); } while (record != NULL); /* @@ -1789,11 +1789,19 @@ PerformWalRecovery(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, - (errmsg("redo is not required"))); + errmsg("redo is not required")); + + } + else + { + /* broken record found */ + ereport(WARNING, + errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database.")); } /* @@ -3024,6 +3032,7 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogPrefetcherReadRecord(xlogprefetcher, &errormsg); if (record == NULL) @@ -3045,6 +3054,18 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * EndRecPtr is the LSN we tried to read but failed. In the + * case of decoding error, it is at the end of the failed + * record but we don't have a means for now to know EndRecPtr + * is pointing to which of the beginning or ending of the + * failed record. + */ + ErrRecPtr = xlogreader->EndRecPtr; } if (readFile >= 0) @@ -3055,13 +3076,15 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, /* * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * failed- in that case we already logged something. In StandbyMode + * that only happens if we have been triggered, so we shouldn't + * loop anymore in that case. When EndOfWAL is true, we don't emit + * the message immediately and instead will show it as a part of a + * decent end-of-wal message later. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), - (errmsg_internal("%s", errormsg) /* already translated */ )); + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), + errmsg_internal("%s", errormsg) /* already translated */ ); } /* @@ -3091,11 +3114,14 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, /* Great, got a record */ return record; } - else - { - /* No valid record available from this source */ - lastSourceFailed = true; + Assert(ErrRecPtr != InvalidXLogRecPtr); + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) + { /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -3108,11 +3134,16 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + errmsg_internal("reached end of WAL at %X/%X on timeline %u in pg_wal during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI)); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -3133,12 +3164,24 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * recovery ended. + * + * Emit a decent message if we met end-of-WAL. Otherwise we should + * have already emitted an error message. + */ + if (xlogreader->EndOfWAL) + ereport(LOG, + errmsg("reached end of WAL at %X/%X on timeline %u", + LSN_FORMAT_ARGS(ErrRecPtr), replayTLI), + (errormsg ? errdetail_internal("%s", errormsg) : 0)); } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -3233,11 +3276,16 @@ retry: case XLREAD_WOULDBLOCK: return XLREAD_WOULDBLOCK; case XLREAD_FAIL: + Assert(!StandbyMode || CheckForStandbyTrigger()); + if (readFile >= 0) close(readFile); readFile = -1; readLen = 0; readSource = XLOG_FROM_ANY; + + /* promotion exit is not end-of-WAL */ + xlogreader->EndOfWAL = !StandbyMode; return XLREAD_FAIL; case XLREAD_SUCCESS: break; @@ -3898,7 +3946,8 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + /* use currentSource as readSource is reset at failure */ + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index 6cbb67c92a3..7ef19e435ce 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -472,10 +472,9 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", - startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); + errmsg("replication terminated by primary server at %X/%X on timeline %u.", + LSN_FORMAT_ARGS(LogstreamResult.Write), + startpointTLI)); endofwal = true; break; } diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index 9993378ca58..26a4125b301 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -1168,9 +1168,16 @@ main(int argc, char **argv) exit(0); if (errormsg) - pg_fatal("error in WAL record at %X/%X: %s", - LSN_FORMAT_ARGS(xlogreader_state->ReadRecPtr), - errormsg); + { + if (xlogreader_state->EndOfWAL) + pg_log_info("end of WAL at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + else + pg_fatal("error in WAL record at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + } XLogReaderFree(xlogreader_state); diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index e87f91316ae..70d3b25edaf 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -205,6 +205,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* was the last attempt EOW? */ /* * Set at the end of recovery: the start point of a partial record at the diff --git a/src/test/recovery/t/011_crash_recovery.pl b/src/test/recovery/t/011_crash_recovery.pl index 1b57d01046d..bde16b7cfa7 100644 --- a/src/test/recovery/t/011_crash_recovery.pl +++ b/src/test/recovery/t/011_crash_recovery.pl @@ -9,7 +9,9 @@ use warnings; use PostgreSQL::Test::Cluster; use PostgreSQL::Test::Utils; use Test::More; +use IPC::Run; +my $reached_eow_pat = "reached end of WAL at "; my $node = PostgreSQL::Test::Cluster->new('primary'); $node->init(allows_streaming => 1); $node->start; @@ -47,7 +49,15 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), # Crash and restart the postmaster $node->stop('immediate'); +my $logstart = get_log_size($node); $node->start; +my $max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, $reached_eow_pat, $logstart)); + sleep 0.5; +} +ok ($max_attempts >= 0, "end-of-wal is logged"); # Make sure we really got a new xid cmp_ok($node->safe_psql('postgres', 'SELECT pg_current_xact_id()'), @@ -60,4 +70,100 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), $stdin .= "\\q\n"; $tx->finish; # wait for psql to quit gracefully +my $segsize = $node->safe_psql('postgres', + qq[SELECT setting FROM pg_settings WHERE name = 'wal_segment_size';]); + +# make sure no records afterwards go to the next segment +$node->safe_psql('postgres', qq[ + SELECT pg_switch_wal(); + CHECKPOINT; + CREATE TABLE t(); +]); +$node->stop('immediate'); + +# identify REDO WAL file +my $cmd = "pg_controldata -D " . $node->data_dir(); +$cmd = ['pg_controldata', '-D', $node->data_dir()]; +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stdout =~ /^Latest checkpoint's REDO WAL file:[ \t] *(.+)$/m, + "checkpoint file is identified"); +my $chkptfile = $1; + +# identify the last record +my $walfile = $node->data_dir() . "/pg_wal/$chkptfile"; +$cmd = ['pg_waldump', $walfile]; +$stdout = ''; +$stderr = ''; +my $lastrec; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +foreach my $l (split(/\r?\n/, $stdout)) +{ + $lastrec = $l; +} +ok(defined $lastrec, "last WAL record is extracted"); +ok($stderr =~ /end of WAL at ([0-9A-F\/]+): .* at \g1/, + "pg_waldump emits the correct ending message"); + +# read the last record LSN excluding leading zeroes +ok ($lastrec =~ /, lsn: 0\/0*([1-9A-F][0-9A-F]+),/, + "LSN of the last record identified"); +my $lastlsn = $1; + +# corrupt the last record +my $offset = hex($lastlsn) % $segsize; +open(my $segf, '+<', $walfile) or die "failed to open $walfile\n"; +seek($segf, $offset, 0); # halfway corrupt the last record +print $segf "\0\0\0\0"; +close($segf); + +# pg_waldump complains about the corrupted record +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stderr =~ /error: error in WAL record at 0\/$lastlsn: .* at 0\/$lastlsn/, + "pg_waldump emits the correct error message"); + +# also server complains +$logstart = get_log_size($node); +$node->start; +$max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, "WARNING: invalid record length at 0/$lastlsn: wanted [0-9]+, got 0", + $logstart)); + sleep 0.5; +} +ok($max_attempts >= 0, "header error is logged at $lastlsn"); + +# no end-of-wal message should be seen this time +ok(!find_in_log($node, $reached_eow_pat, $logstart), + "false log message is not emitted"); + +$node->stop('immediate'); + done_testing(); + +#### helper routines +# return the size of logfile of $node in bytes +sub get_log_size +{ + my ($node) = @_; + + return (stat $node->logfile)[7]; +} + +# find $pat in logfile of $node after $off-th byte +sub find_in_log +{ + my ($node, $pat, $off) = @_; + + $off = 0 unless defined $off; + my $log = PostgreSQL::Test::Utils::slurp_file($node->logfile); + return 0 if (length($log) <= $off); + + $log = substr($log, $off); + + return $log =~ m/$pat/; +} -- 2.25.1 --nHJAUhyIZkPvF01C-- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH] Make End-Of-Recovery error less scary @ 2022-07-07 02:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2022-07-07 02:51 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlogreader.c | 134 +++++++++++++++++----- src/backend/access/transam/xlogrecovery.c | 95 +++++++++++---- src/backend/replication/walreceiver.c | 7 +- src/bin/pg_waldump/pg_waldump.c | 13 ++- src/include/access/xlogreader.h | 1 + src/test/recovery/t/011_crash_recovery.pl | 106 +++++++++++++++++ 6 files changed, 298 insertions(+), 58 deletions(-) diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 050d2f424e4..9b8f29d0ad0 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -48,6 +48,8 @@ static int ReadPageInternal(XLogReaderState *state, XLogRecPtr pageptr, int reqLen); static void XLogReaderInvalReadState(XLogReaderState *state); static XLogPageReadResult XLogDecodeNextRecord(XLogReaderState *state, bool non_blocking); +static bool ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record); static bool ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess); static bool ValidXLogRecord(XLogReaderState *state, XLogRecord *record, @@ -149,6 +151,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -558,6 +561,7 @@ XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking) /* reset error state */ state->errormsg_buf[0] = '\0'; decoded = NULL; + state->EndOfWAL = false; state->abortedRecPtr = InvalidXLogRecPtr; state->missingContrecPtr = InvalidXLogRecPtr; @@ -640,25 +644,21 @@ restart: Assert(pageHeaderSize <= readOff); /* - * Read the record length. + * Validate the record header. * - * NB: Even though we use an XLogRecord pointer here, the whole record - * header might not fit on this page. xl_tot_len is the first field of the - * struct, so it must be on this page (the records are MAXALIGNed), but we - * cannot access any other fields until we've verified that we got the - * whole header. - */ - record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); - total_len = record->xl_tot_len; - - /* - * If the whole record header is on this page, validate it immediately. - * Otherwise do just a basic sanity check on xl_tot_len, and validate the - * rest of the header after reading it from the next page. The xl_tot_len + * Even though we use an XLogRecord pointer here, the whole record header + * might not fit on this page. If the whole record header is on this page, + * validate it immediately. Even otherwise xl_tot_len must be on this page + * (it is the first field of MAXALIGNed records), but we still cannot + * access any further fields until we've verified that we got the whole + * header, so do just a basic sanity check on record length, and validate + * the rest of the header after reading it from the next page. The length * check is necessary here to ensure that we enter the "Need to reassemble * record" code path below; otherwise we might fail to apply * ValidXLogRecordHeader at all. */ + record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); + if (targetRecOff <= XLOG_BLCKSZ - SizeOfXLogRecord) { if (!ValidXLogRecordHeader(state, RecPtr, state->DecodeRecPtr, record, @@ -668,18 +668,14 @@ restart: } else { - /* XXX: more validation should be done here */ - if (total_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: wanted %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, total_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) goto err; - } + gotheader = false; } + total_len = record->xl_tot_len; + /* * Find space to decode this record. Don't allow oversized allocation if * the caller requested nonblocking. Otherwise, we *have* to try to @@ -1106,16 +1102,47 @@ XLogReaderInvalReadState(XLogReaderState *state) } /* - * Validate an XLOG record header. + * Validate record length of an XLOG record header. * - * This is just a convenience subroutine to avoid duplicated code in - * XLogReadRecord. It's not intended for use from anywhere else. + * This is substantially a part of ValidXLogRecordHeader. But XLogReadRecord + * needs this separate from the function in case of a partial record header. */ static bool -ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, - XLogRecPtr PrevRecPtr, XLogRecord *record, - bool randAccess) +ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record) { + if (record->xl_tot_len == 0) + { + char *p; + char *pe; + + /* + * We are almost sure reaching the end of WAL, make sure that the + * whole page after the record is filled with zeroes. + */ + p = (char *) record; + pe = p + XLOG_BLCKSZ - (RecPtr & (XLOG_BLCKSZ - 1)); + + while (*p == 0 && p < pe) + p++; + + if (p == pe) + { + /* + * The page after the record is completely zeroed. That suggests + * we don't have a record after this point. We don't bother + * checking the pages after since they are not zeroed in the case + * of recycled segments. + */ + report_invalid_record(state, "empty record at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + return false; + } + } + if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, @@ -1124,6 +1151,24 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, (uint32) SizeOfXLogRecord, record->xl_tot_len); return false; } + + return true; +} + +/* + * Validate an XLOG record header. + * + * This is just a convenience subroutine to avoid duplicated code in + * XLogReadRecord. It's not intended for use from anywhere else. + */ +static bool +ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecPtr PrevRecPtr, XLogRecord *record, + bool randAccess) +{ + if (!ValidXLogRecordLength(state, RecPtr, record)) + return false; + if (!RmgrIdIsValid(record->xl_rmid)) { report_invalid_record(state, @@ -1222,6 +1267,31 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, XLogSegNoOffsetToRecPtr(segno, offset, state->segcxt.ws_segsize, recaddr); + StaticAssertStmt(XLOG_PAGE_MAGIC != 0, "XLOG_PAGE_MAGIC is zero"); + + if (hdr->xlp_magic == 0) + { + /* Regard an empty page as End-Of-WAL */ + int i; + + for (i = 0; i < XLOG_BLCKSZ && phdr[i] == 0; i++); + if (i == XLOG_BLCKSZ) + { + char fname[MAXFNAMELEN]; + + XLogFileName(fname, state->seg.ws_tli, segno, + state->segcxt.ws_segsize); + + report_invalid_record(state, + "empty page in log segment %s, offset %u", + fname, + offset); + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid magic number */ + } if (hdr->xlp_magic != XLOG_PAGE_MAGIC) { char fname[MAXFNAMELEN]; @@ -1307,6 +1377,14 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, LSN_FORMAT_ARGS(hdr->xlp_pageaddr), fname, offset); + + /* + * If the page address is less than expected we assume it is an unused + * page in a recycled segment. + */ + if (hdr->xlp_pageaddr < recaddr) + state->EndOfWAL = true; + return false; } diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index b41e6826643..caa3b5e5b31 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -1626,7 +1626,7 @@ PerformWalRecovery(void) /* just have to read next record after CheckPoint */ Assert(xlogreader->ReadRecPtr == CheckPointLoc); replayTLI = CheckPointTLI; - record = ReadRecord(xlogprefetcher, LOG, false, replayTLI); + record = ReadRecord(xlogprefetcher, WARNING, false, replayTLI); } if (record != NULL) @@ -1735,7 +1735,7 @@ PerformWalRecovery(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogprefetcher, LOG, false, replayTLI); + record = ReadRecord(xlogprefetcher, WARNING, false, replayTLI); } while (record != NULL); /* @@ -1789,11 +1789,19 @@ PerformWalRecovery(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, - (errmsg("redo is not required"))); + errmsg("redo is not required")); + + } + else + { + /* broken record found */ + ereport(WARNING, + errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database.")); } /* @@ -3024,6 +3032,7 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogPrefetcherReadRecord(xlogprefetcher, &errormsg); if (record == NULL) @@ -3045,6 +3054,18 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * EndRecPtr is the LSN we tried to read but failed. In the + * case of decoding error, it is at the end of the failed + * record but we don't have a means for now to know EndRecPtr + * is pointing to which of the beginning or ending of the + * failed record. + */ + ErrRecPtr = xlogreader->EndRecPtr; } if (readFile >= 0) @@ -3055,13 +3076,15 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, /* * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * failed- in that case we already logged something. In StandbyMode + * that only happens if we have been triggered, so we shouldn't + * loop anymore in that case. When EndOfWAL is true, we don't emit + * the message immediately and instead will show it as a part of a + * decent end-of-wal message later. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), - (errmsg_internal("%s", errormsg) /* already translated */ )); + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), + errmsg_internal("%s", errormsg) /* already translated */ ); } /* @@ -3091,11 +3114,14 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, /* Great, got a record */ return record; } - else - { - /* No valid record available from this source */ - lastSourceFailed = true; + Assert(ErrRecPtr != InvalidXLogRecPtr); + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) + { /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -3108,11 +3134,16 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + errmsg_internal("reached end of WAL at %X/%X on timeline %u in pg_wal during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI)); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -3133,12 +3164,24 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * recovery ended. + * + * Emit a decent message if we met end-of-WAL. Otherwise we should + * have already emitted an error message. + */ + if (xlogreader->EndOfWAL) + ereport(LOG, + errmsg("reached end of WAL at %X/%X on timeline %u", + LSN_FORMAT_ARGS(ErrRecPtr), replayTLI), + (errormsg ? errdetail_internal("%s", errormsg) : 0)); } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -3233,11 +3276,16 @@ retry: case XLREAD_WOULDBLOCK: return XLREAD_WOULDBLOCK; case XLREAD_FAIL: + Assert(!StandbyMode || CheckForStandbyTrigger()); + if (readFile >= 0) close(readFile); readFile = -1; readLen = 0; readSource = XLOG_FROM_ANY; + + /* promotion exit is not end-of-WAL */ + xlogreader->EndOfWAL = !StandbyMode; return XLREAD_FAIL; case XLREAD_SUCCESS: break; @@ -3898,7 +3946,8 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + /* use currentSource as readSource is reset at failure */ + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index f6ef0ace2c4..c6d7be66885 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -472,10 +472,9 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", - startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); + errmsg("replication terminated by primary server at %X/%X on timeline %u.", + LSN_FORMAT_ARGS(LogstreamResult.Write), + startpointTLI)); endofwal = true; break; } diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index 9993378ca58..26a4125b301 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -1168,9 +1168,16 @@ main(int argc, char **argv) exit(0); if (errormsg) - pg_fatal("error in WAL record at %X/%X: %s", - LSN_FORMAT_ARGS(xlogreader_state->ReadRecPtr), - errormsg); + { + if (xlogreader_state->EndOfWAL) + pg_log_info("end of WAL at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + else + pg_fatal("error in WAL record at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + } XLogReaderFree(xlogreader_state); diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 6afec33d418..264afb6a78e 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -205,6 +205,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* was the last attempt EOW? */ /* * Set at the end of recovery: the start point of a partial record at the diff --git a/src/test/recovery/t/011_crash_recovery.pl b/src/test/recovery/t/011_crash_recovery.pl index 1b57d01046d..bde16b7cfa7 100644 --- a/src/test/recovery/t/011_crash_recovery.pl +++ b/src/test/recovery/t/011_crash_recovery.pl @@ -9,7 +9,9 @@ use warnings; use PostgreSQL::Test::Cluster; use PostgreSQL::Test::Utils; use Test::More; +use IPC::Run; +my $reached_eow_pat = "reached end of WAL at "; my $node = PostgreSQL::Test::Cluster->new('primary'); $node->init(allows_streaming => 1); $node->start; @@ -47,7 +49,15 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), # Crash and restart the postmaster $node->stop('immediate'); +my $logstart = get_log_size($node); $node->start; +my $max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, $reached_eow_pat, $logstart)); + sleep 0.5; +} +ok ($max_attempts >= 0, "end-of-wal is logged"); # Make sure we really got a new xid cmp_ok($node->safe_psql('postgres', 'SELECT pg_current_xact_id()'), @@ -60,4 +70,100 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), $stdin .= "\\q\n"; $tx->finish; # wait for psql to quit gracefully +my $segsize = $node->safe_psql('postgres', + qq[SELECT setting FROM pg_settings WHERE name = 'wal_segment_size';]); + +# make sure no records afterwards go to the next segment +$node->safe_psql('postgres', qq[ + SELECT pg_switch_wal(); + CHECKPOINT; + CREATE TABLE t(); +]); +$node->stop('immediate'); + +# identify REDO WAL file +my $cmd = "pg_controldata -D " . $node->data_dir(); +$cmd = ['pg_controldata', '-D', $node->data_dir()]; +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stdout =~ /^Latest checkpoint's REDO WAL file:[ \t] *(.+)$/m, + "checkpoint file is identified"); +my $chkptfile = $1; + +# identify the last record +my $walfile = $node->data_dir() . "/pg_wal/$chkptfile"; +$cmd = ['pg_waldump', $walfile]; +$stdout = ''; +$stderr = ''; +my $lastrec; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +foreach my $l (split(/\r?\n/, $stdout)) +{ + $lastrec = $l; +} +ok(defined $lastrec, "last WAL record is extracted"); +ok($stderr =~ /end of WAL at ([0-9A-F\/]+): .* at \g1/, + "pg_waldump emits the correct ending message"); + +# read the last record LSN excluding leading zeroes +ok ($lastrec =~ /, lsn: 0\/0*([1-9A-F][0-9A-F]+),/, + "LSN of the last record identified"); +my $lastlsn = $1; + +# corrupt the last record +my $offset = hex($lastlsn) % $segsize; +open(my $segf, '+<', $walfile) or die "failed to open $walfile\n"; +seek($segf, $offset, 0); # halfway corrupt the last record +print $segf "\0\0\0\0"; +close($segf); + +# pg_waldump complains about the corrupted record +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stderr =~ /error: error in WAL record at 0\/$lastlsn: .* at 0\/$lastlsn/, + "pg_waldump emits the correct error message"); + +# also server complains +$logstart = get_log_size($node); +$node->start; +$max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, "WARNING: invalid record length at 0/$lastlsn: wanted [0-9]+, got 0", + $logstart)); + sleep 0.5; +} +ok($max_attempts >= 0, "header error is logged at $lastlsn"); + +# no end-of-wal message should be seen this time +ok(!find_in_log($node, $reached_eow_pat, $logstart), + "false log message is not emitted"); + +$node->stop('immediate'); + done_testing(); + +#### helper routines +# return the size of logfile of $node in bytes +sub get_log_size +{ + my ($node) = @_; + + return (stat $node->logfile)[7]; +} + +# find $pat in logfile of $node after $off-th byte +sub find_in_log +{ + my ($node, $pat, $off) = @_; + + $off = 0 unless defined $off; + my $log = PostgreSQL::Test::Utils::slurp_file($node->logfile); + return 0 if (length($log) <= $off); + + $log = substr($log, $off); + + return $log =~ m/$pat/; +} -- 2.25.1 --sMP3l7aO1Ldl3x+Z-- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH] Make End-Of-Recovery error less scary @ 2022-07-07 02:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2022-07-07 02:51 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlogreader.c | 134 +++++++++++++++++----- src/backend/access/transam/xlogrecovery.c | 95 +++++++++++---- src/backend/replication/walreceiver.c | 7 +- src/bin/pg_waldump/pg_waldump.c | 13 ++- src/include/access/xlogreader.h | 1 + src/test/recovery/t/011_crash_recovery.pl | 106 +++++++++++++++++ 6 files changed, 298 insertions(+), 58 deletions(-) diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 050d2f424e4..9b8f29d0ad0 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -48,6 +48,8 @@ static int ReadPageInternal(XLogReaderState *state, XLogRecPtr pageptr, int reqLen); static void XLogReaderInvalReadState(XLogReaderState *state); static XLogPageReadResult XLogDecodeNextRecord(XLogReaderState *state, bool non_blocking); +static bool ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record); static bool ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess); static bool ValidXLogRecord(XLogReaderState *state, XLogRecord *record, @@ -149,6 +151,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -558,6 +561,7 @@ XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking) /* reset error state */ state->errormsg_buf[0] = '\0'; decoded = NULL; + state->EndOfWAL = false; state->abortedRecPtr = InvalidXLogRecPtr; state->missingContrecPtr = InvalidXLogRecPtr; @@ -640,25 +644,21 @@ restart: Assert(pageHeaderSize <= readOff); /* - * Read the record length. + * Validate the record header. * - * NB: Even though we use an XLogRecord pointer here, the whole record - * header might not fit on this page. xl_tot_len is the first field of the - * struct, so it must be on this page (the records are MAXALIGNed), but we - * cannot access any other fields until we've verified that we got the - * whole header. - */ - record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); - total_len = record->xl_tot_len; - - /* - * If the whole record header is on this page, validate it immediately. - * Otherwise do just a basic sanity check on xl_tot_len, and validate the - * rest of the header after reading it from the next page. The xl_tot_len + * Even though we use an XLogRecord pointer here, the whole record header + * might not fit on this page. If the whole record header is on this page, + * validate it immediately. Even otherwise xl_tot_len must be on this page + * (it is the first field of MAXALIGNed records), but we still cannot + * access any further fields until we've verified that we got the whole + * header, so do just a basic sanity check on record length, and validate + * the rest of the header after reading it from the next page. The length * check is necessary here to ensure that we enter the "Need to reassemble * record" code path below; otherwise we might fail to apply * ValidXLogRecordHeader at all. */ + record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); + if (targetRecOff <= XLOG_BLCKSZ - SizeOfXLogRecord) { if (!ValidXLogRecordHeader(state, RecPtr, state->DecodeRecPtr, record, @@ -668,18 +668,14 @@ restart: } else { - /* XXX: more validation should be done here */ - if (total_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: wanted %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, total_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) goto err; - } + gotheader = false; } + total_len = record->xl_tot_len; + /* * Find space to decode this record. Don't allow oversized allocation if * the caller requested nonblocking. Otherwise, we *have* to try to @@ -1106,16 +1102,47 @@ XLogReaderInvalReadState(XLogReaderState *state) } /* - * Validate an XLOG record header. + * Validate record length of an XLOG record header. * - * This is just a convenience subroutine to avoid duplicated code in - * XLogReadRecord. It's not intended for use from anywhere else. + * This is substantially a part of ValidXLogRecordHeader. But XLogReadRecord + * needs this separate from the function in case of a partial record header. */ static bool -ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, - XLogRecPtr PrevRecPtr, XLogRecord *record, - bool randAccess) +ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record) { + if (record->xl_tot_len == 0) + { + char *p; + char *pe; + + /* + * We are almost sure reaching the end of WAL, make sure that the + * whole page after the record is filled with zeroes. + */ + p = (char *) record; + pe = p + XLOG_BLCKSZ - (RecPtr & (XLOG_BLCKSZ - 1)); + + while (*p == 0 && p < pe) + p++; + + if (p == pe) + { + /* + * The page after the record is completely zeroed. That suggests + * we don't have a record after this point. We don't bother + * checking the pages after since they are not zeroed in the case + * of recycled segments. + */ + report_invalid_record(state, "empty record at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + return false; + } + } + if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, @@ -1124,6 +1151,24 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, (uint32) SizeOfXLogRecord, record->xl_tot_len); return false; } + + return true; +} + +/* + * Validate an XLOG record header. + * + * This is just a convenience subroutine to avoid duplicated code in + * XLogReadRecord. It's not intended for use from anywhere else. + */ +static bool +ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecPtr PrevRecPtr, XLogRecord *record, + bool randAccess) +{ + if (!ValidXLogRecordLength(state, RecPtr, record)) + return false; + if (!RmgrIdIsValid(record->xl_rmid)) { report_invalid_record(state, @@ -1222,6 +1267,31 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, XLogSegNoOffsetToRecPtr(segno, offset, state->segcxt.ws_segsize, recaddr); + StaticAssertStmt(XLOG_PAGE_MAGIC != 0, "XLOG_PAGE_MAGIC is zero"); + + if (hdr->xlp_magic == 0) + { + /* Regard an empty page as End-Of-WAL */ + int i; + + for (i = 0; i < XLOG_BLCKSZ && phdr[i] == 0; i++); + if (i == XLOG_BLCKSZ) + { + char fname[MAXFNAMELEN]; + + XLogFileName(fname, state->seg.ws_tli, segno, + state->segcxt.ws_segsize); + + report_invalid_record(state, + "empty page in log segment %s, offset %u", + fname, + offset); + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid magic number */ + } if (hdr->xlp_magic != XLOG_PAGE_MAGIC) { char fname[MAXFNAMELEN]; @@ -1307,6 +1377,14 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, LSN_FORMAT_ARGS(hdr->xlp_pageaddr), fname, offset); + + /* + * If the page address is less than expected we assume it is an unused + * page in a recycled segment. + */ + if (hdr->xlp_pageaddr < recaddr) + state->EndOfWAL = true; + return false; } diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index b41e6826643..caa3b5e5b31 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -1626,7 +1626,7 @@ PerformWalRecovery(void) /* just have to read next record after CheckPoint */ Assert(xlogreader->ReadRecPtr == CheckPointLoc); replayTLI = CheckPointTLI; - record = ReadRecord(xlogprefetcher, LOG, false, replayTLI); + record = ReadRecord(xlogprefetcher, WARNING, false, replayTLI); } if (record != NULL) @@ -1735,7 +1735,7 @@ PerformWalRecovery(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogprefetcher, LOG, false, replayTLI); + record = ReadRecord(xlogprefetcher, WARNING, false, replayTLI); } while (record != NULL); /* @@ -1789,11 +1789,19 @@ PerformWalRecovery(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, - (errmsg("redo is not required"))); + errmsg("redo is not required")); + + } + else + { + /* broken record found */ + ereport(WARNING, + errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database.")); } /* @@ -3024,6 +3032,7 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogPrefetcherReadRecord(xlogprefetcher, &errormsg); if (record == NULL) @@ -3045,6 +3054,18 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * EndRecPtr is the LSN we tried to read but failed. In the + * case of decoding error, it is at the end of the failed + * record but we don't have a means for now to know EndRecPtr + * is pointing to which of the beginning or ending of the + * failed record. + */ + ErrRecPtr = xlogreader->EndRecPtr; } if (readFile >= 0) @@ -3055,13 +3076,15 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, /* * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * failed- in that case we already logged something. In StandbyMode + * that only happens if we have been triggered, so we shouldn't + * loop anymore in that case. When EndOfWAL is true, we don't emit + * the message immediately and instead will show it as a part of a + * decent end-of-wal message later. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), - (errmsg_internal("%s", errormsg) /* already translated */ )); + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), + errmsg_internal("%s", errormsg) /* already translated */ ); } /* @@ -3091,11 +3114,14 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, /* Great, got a record */ return record; } - else - { - /* No valid record available from this source */ - lastSourceFailed = true; + Assert(ErrRecPtr != InvalidXLogRecPtr); + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) + { /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -3108,11 +3134,16 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + errmsg_internal("reached end of WAL at %X/%X on timeline %u in pg_wal during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI)); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -3133,12 +3164,24 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * recovery ended. + * + * Emit a decent message if we met end-of-WAL. Otherwise we should + * have already emitted an error message. + */ + if (xlogreader->EndOfWAL) + ereport(LOG, + errmsg("reached end of WAL at %X/%X on timeline %u", + LSN_FORMAT_ARGS(ErrRecPtr), replayTLI), + (errormsg ? errdetail_internal("%s", errormsg) : 0)); } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -3233,11 +3276,16 @@ retry: case XLREAD_WOULDBLOCK: return XLREAD_WOULDBLOCK; case XLREAD_FAIL: + Assert(!StandbyMode || CheckForStandbyTrigger()); + if (readFile >= 0) close(readFile); readFile = -1; readLen = 0; readSource = XLOG_FROM_ANY; + + /* promotion exit is not end-of-WAL */ + xlogreader->EndOfWAL = !StandbyMode; return XLREAD_FAIL; case XLREAD_SUCCESS: break; @@ -3898,7 +3946,8 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + /* use currentSource as readSource is reset at failure */ + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index f6ef0ace2c4..c6d7be66885 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -472,10 +472,9 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", - startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); + errmsg("replication terminated by primary server at %X/%X on timeline %u.", + LSN_FORMAT_ARGS(LogstreamResult.Write), + startpointTLI)); endofwal = true; break; } diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index 9993378ca58..26a4125b301 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -1168,9 +1168,16 @@ main(int argc, char **argv) exit(0); if (errormsg) - pg_fatal("error in WAL record at %X/%X: %s", - LSN_FORMAT_ARGS(xlogreader_state->ReadRecPtr), - errormsg); + { + if (xlogreader_state->EndOfWAL) + pg_log_info("end of WAL at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + else + pg_fatal("error in WAL record at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + } XLogReaderFree(xlogreader_state); diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index 6afec33d418..264afb6a78e 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -205,6 +205,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* was the last attempt EOW? */ /* * Set at the end of recovery: the start point of a partial record at the diff --git a/src/test/recovery/t/011_crash_recovery.pl b/src/test/recovery/t/011_crash_recovery.pl index 1b57d01046d..bde16b7cfa7 100644 --- a/src/test/recovery/t/011_crash_recovery.pl +++ b/src/test/recovery/t/011_crash_recovery.pl @@ -9,7 +9,9 @@ use warnings; use PostgreSQL::Test::Cluster; use PostgreSQL::Test::Utils; use Test::More; +use IPC::Run; +my $reached_eow_pat = "reached end of WAL at "; my $node = PostgreSQL::Test::Cluster->new('primary'); $node->init(allows_streaming => 1); $node->start; @@ -47,7 +49,15 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), # Crash and restart the postmaster $node->stop('immediate'); +my $logstart = get_log_size($node); $node->start; +my $max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, $reached_eow_pat, $logstart)); + sleep 0.5; +} +ok ($max_attempts >= 0, "end-of-wal is logged"); # Make sure we really got a new xid cmp_ok($node->safe_psql('postgres', 'SELECT pg_current_xact_id()'), @@ -60,4 +70,100 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), $stdin .= "\\q\n"; $tx->finish; # wait for psql to quit gracefully +my $segsize = $node->safe_psql('postgres', + qq[SELECT setting FROM pg_settings WHERE name = 'wal_segment_size';]); + +# make sure no records afterwards go to the next segment +$node->safe_psql('postgres', qq[ + SELECT pg_switch_wal(); + CHECKPOINT; + CREATE TABLE t(); +]); +$node->stop('immediate'); + +# identify REDO WAL file +my $cmd = "pg_controldata -D " . $node->data_dir(); +$cmd = ['pg_controldata', '-D', $node->data_dir()]; +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stdout =~ /^Latest checkpoint's REDO WAL file:[ \t] *(.+)$/m, + "checkpoint file is identified"); +my $chkptfile = $1; + +# identify the last record +my $walfile = $node->data_dir() . "/pg_wal/$chkptfile"; +$cmd = ['pg_waldump', $walfile]; +$stdout = ''; +$stderr = ''; +my $lastrec; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +foreach my $l (split(/\r?\n/, $stdout)) +{ + $lastrec = $l; +} +ok(defined $lastrec, "last WAL record is extracted"); +ok($stderr =~ /end of WAL at ([0-9A-F\/]+): .* at \g1/, + "pg_waldump emits the correct ending message"); + +# read the last record LSN excluding leading zeroes +ok ($lastrec =~ /, lsn: 0\/0*([1-9A-F][0-9A-F]+),/, + "LSN of the last record identified"); +my $lastlsn = $1; + +# corrupt the last record +my $offset = hex($lastlsn) % $segsize; +open(my $segf, '+<', $walfile) or die "failed to open $walfile\n"; +seek($segf, $offset, 0); # halfway corrupt the last record +print $segf "\0\0\0\0"; +close($segf); + +# pg_waldump complains about the corrupted record +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stderr =~ /error: error in WAL record at 0\/$lastlsn: .* at 0\/$lastlsn/, + "pg_waldump emits the correct error message"); + +# also server complains +$logstart = get_log_size($node); +$node->start; +$max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, "WARNING: invalid record length at 0/$lastlsn: wanted [0-9]+, got 0", + $logstart)); + sleep 0.5; +} +ok($max_attempts >= 0, "header error is logged at $lastlsn"); + +# no end-of-wal message should be seen this time +ok(!find_in_log($node, $reached_eow_pat, $logstart), + "false log message is not emitted"); + +$node->stop('immediate'); + done_testing(); + +#### helper routines +# return the size of logfile of $node in bytes +sub get_log_size +{ + my ($node) = @_; + + return (stat $node->logfile)[7]; +} + +# find $pat in logfile of $node after $off-th byte +sub find_in_log +{ + my ($node, $pat, $off) = @_; + + $off = 0 unless defined $off; + my $log = PostgreSQL::Test::Utils::slurp_file($node->logfile); + return 0 if (length($log) <= $off); + + $log = substr($log, $off); + + return $log =~ m/$pat/; +} -- 2.25.1 --sMP3l7aO1Ldl3x+Z-- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH] Make End-Of-Recovery error less scary @ 2022-11-15 04:41 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2022-11-15 04:41 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlogreader.c | 135 +++++++++++++++++----- src/backend/access/transam/xlogrecovery.c | 94 +++++++++++---- src/backend/replication/walreceiver.c | 7 +- src/bin/pg_waldump/pg_waldump.c | 13 ++- src/include/access/xlogreader.h | 1 + src/test/recovery/t/011_crash_recovery.pl | 106 +++++++++++++++++ 6 files changed, 298 insertions(+), 58 deletions(-) diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 93f667b2544..137de967951 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -48,6 +48,8 @@ static int ReadPageInternal(XLogReaderState *state, XLogRecPtr pageptr, int reqLen); static void XLogReaderInvalReadState(XLogReaderState *state); static XLogPageReadResult XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking); +static bool ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record); static bool ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess); static bool ValidXLogRecord(XLogReaderState *state, XLogRecord *record, @@ -149,6 +151,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -558,6 +561,7 @@ XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking) /* reset error state */ state->errormsg_buf[0] = '\0'; decoded = NULL; + state->EndOfWAL = false; state->abortedRecPtr = InvalidXLogRecPtr; state->missingContrecPtr = InvalidXLogRecPtr; @@ -640,25 +644,21 @@ restart: Assert(pageHeaderSize <= readOff); /* - * Read the record length. + * Validate the record header. * - * NB: Even though we use an XLogRecord pointer here, the whole record - * header might not fit on this page. xl_tot_len is the first field of the - * struct, so it must be on this page (the records are MAXALIGNed), but we - * cannot access any other fields until we've verified that we got the - * whole header. - */ - record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); - total_len = record->xl_tot_len; - - /* - * If the whole record header is on this page, validate it immediately. - * Otherwise do just a basic sanity check on xl_tot_len, and validate the - * rest of the header after reading it from the next page. The xl_tot_len + * Even though we use an XLogRecord pointer here, the whole record header + * might not fit on this page. If the whole record header is on this page, + * validate it immediately. Even otherwise xl_tot_len must be on this page + * (it is the first field of MAXALIGNed records), but we still cannot + * access any further fields until we've verified that we got the whole + * header, so do just a basic sanity check on record length, and validate + * the rest of the header after reading it from the next page. The length * check is necessary here to ensure that we enter the "Need to reassemble * record" code path below; otherwise we might fail to apply * ValidXLogRecordHeader at all. */ + record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); + if (targetRecOff <= XLOG_BLCKSZ - SizeOfXLogRecord) { if (!ValidXLogRecordHeader(state, RecPtr, state->DecodeRecPtr, record, @@ -668,18 +668,14 @@ restart: } else { - /* XXX: more validation should be done here */ - if (total_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: wanted %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, total_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) goto err; - } + gotheader = false; } + total_len = record->xl_tot_len; + /* * Find space to decode this record. Don't allow oversized allocation if * the caller requested nonblocking. Otherwise, we *have* to try to @@ -1106,16 +1102,47 @@ XLogReaderInvalReadState(XLogReaderState *state) } /* - * Validate an XLOG record header. + * Validate record length of an XLOG record header. * - * This is just a convenience subroutine to avoid duplicated code in - * XLogReadRecord. It's not intended for use from anywhere else. + * This is substantially a part of ValidXLogRecordHeader. But XLogReadRecord + * needs this separate from the function in case of a partial record header. */ static bool -ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, - XLogRecPtr PrevRecPtr, XLogRecord *record, - bool randAccess) +ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record) { + if (record->xl_tot_len == 0) + { + char *p; + char *pe; + + /* + * We are almost sure reaching the end of WAL, make sure that the + * whole page after the record is filled with zeroes. + */ + p = (char *) record; + pe = p + XLOG_BLCKSZ - (RecPtr & (XLOG_BLCKSZ - 1)); + + while (p < pe && *p == 0) + p++; + + if (p == pe) + { + /* + * The page after the record is completely zeroed. That suggests + * we don't have a record after this point. We don't bother + * checking the pages after since they are not zeroed in the case + * of recycled segments. + */ + report_invalid_record(state, "empty record at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + return false; + } + } + if (record->xl_tot_len < SizeOfXLogRecord) { report_invalid_record(state, @@ -1124,6 +1151,24 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, (uint32) SizeOfXLogRecord, record->xl_tot_len); return false; } + + return true; +} + +/* + * Validate an XLOG record header. + * + * This is just a convenience subroutine to avoid duplicated code in + * XLogReadRecord. It's not intended for use from anywhere else. + */ +static bool +ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecPtr PrevRecPtr, XLogRecord *record, + bool randAccess) +{ + if (!ValidXLogRecordLength(state, RecPtr, record)) + return false; + if (!RmgrIdIsValid(record->xl_rmid)) { report_invalid_record(state, @@ -1219,6 +1264,32 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, XLByteToSeg(recptr, segno, state->segcxt.ws_segsize); offset = XLogSegmentOffset(recptr, state->segcxt.ws_segsize); + StaticAssertStmt(XLOG_PAGE_MAGIC != 0, "XLOG_PAGE_MAGIC is zero"); + + if (hdr->xlp_magic == 0) + { + /* Regard an empty page as End-Of-WAL */ + int i; + + for (i = 0; i < XLOG_BLCKSZ && phdr[i] == 0; i++); + if (i == XLOG_BLCKSZ) + { + char fname[MAXFNAMELEN]; + + XLogFileName(fname, state->seg.ws_tli, segno, + state->segcxt.ws_segsize); + + report_invalid_record(state, + "empty page in log segment %s, offset %u", + fname, + offset); + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid magic number */ + } + if (hdr->xlp_magic != XLOG_PAGE_MAGIC) { char fname[MAXFNAMELEN]; @@ -1304,6 +1375,14 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, LSN_FORMAT_ARGS(hdr->xlp_pageaddr), fname, offset); + + /* + * If the page address is less than expected we assume it is an unused + * page in a recycled segment. + */ + if (hdr->xlp_pageaddr < recptr) + state->EndOfWAL = true; + return false; } diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index cb07694aea6..3f54c875e5a 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -1626,7 +1626,7 @@ PerformWalRecovery(void) /* just have to read next record after CheckPoint */ Assert(xlogreader->ReadRecPtr == CheckPointLoc); replayTLI = CheckPointTLI; - record = ReadRecord(xlogprefetcher, LOG, false, replayTLI); + record = ReadRecord(xlogprefetcher, WARNING, false, replayTLI); } if (record != NULL) @@ -1735,7 +1735,7 @@ PerformWalRecovery(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogprefetcher, LOG, false, replayTLI); + record = ReadRecord(xlogprefetcher, WARNING, false, replayTLI); } while (record != NULL); /* @@ -1789,11 +1789,18 @@ PerformWalRecovery(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, - (errmsg("redo is not required"))); + errmsg("redo is not required")); + } + else + { + /* broken record found */ + ereport(WARNING, + errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database.")); } /* @@ -3024,6 +3031,7 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogPrefetcherReadRecord(xlogprefetcher, &errormsg); if (record == NULL) @@ -3045,6 +3053,18 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * EndRecPtr is the LSN we tried to read but failed. In the + * case of decoding error, it is at the end of the failed + * record but we don't have a means for now to know EndRecPtr + * is pointing to which of the beginning or ending of the + * failed record. + */ + ErrRecPtr = xlogreader->EndRecPtr; } if (readFile >= 0) @@ -3055,13 +3075,15 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, /* * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * failed- in that case we already logged something. In StandbyMode + * that only happens if we have been triggered, so we shouldn't + * loop anymore in that case. When EndOfWAL is true, we don't emit + * the message immediately and instead will show it as a part of a + * decent end-of-wal message later. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), - (errmsg_internal("%s", errormsg) /* already translated */ )); + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), + errmsg_internal("%s", errormsg) /* already translated */ ); } /* @@ -3091,11 +3113,14 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, /* Great, got a record */ return record; } - else - { - /* No valid record available from this source */ - lastSourceFailed = true; + Assert(ErrRecPtr != InvalidXLogRecPtr); + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) + { /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -3108,11 +3133,16 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + errmsg_internal("reached end of WAL at %X/%X on timeline %u in pg_wal during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI)); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -3133,12 +3163,24 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * recovery ended. + * + * Emit a decent message if we met end-of-WAL. Otherwise we should + * have already emitted an error message. + */ + if (xlogreader->EndOfWAL) + ereport(LOG, + errmsg("reached end of WAL at %X/%X on timeline %u", + LSN_FORMAT_ARGS(ErrRecPtr), replayTLI), + (errormsg ? errdetail_internal("%s", errormsg) : 0)); } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -3233,11 +3275,16 @@ retry: case XLREAD_WOULDBLOCK: return XLREAD_WOULDBLOCK; case XLREAD_FAIL: + Assert(!StandbyMode || CheckForStandbyTrigger()); + if (readFile >= 0) close(readFile); readFile = -1; readLen = 0; readSource = XLOG_FROM_ANY; + + /* promotion exit is not end-of-WAL */ + xlogreader->EndOfWAL = !StandbyMode; return XLREAD_FAIL; case XLREAD_SUCCESS: break; @@ -3898,7 +3945,8 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + /* use currentSource as readSource is reset at failure */ + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index ad383dbcaa6..054a4cb127a 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -498,10 +498,9 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", - startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); + errmsg("replication terminated by primary server at %X/%X on timeline %u.", + LSN_FORMAT_ARGS(LogstreamResult.Write), + startpointTLI)); endofwal = true; break; } diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index 9993378ca58..26a4125b301 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -1168,9 +1168,16 @@ main(int argc, char **argv) exit(0); if (errormsg) - pg_fatal("error in WAL record at %X/%X: %s", - LSN_FORMAT_ARGS(xlogreader_state->ReadRecPtr), - errormsg); + { + if (xlogreader_state->EndOfWAL) + pg_log_info("end of WAL at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + else + pg_fatal("error in WAL record at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + } XLogReaderFree(xlogreader_state); diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index e87f91316ae..70d3b25edaf 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -205,6 +205,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* was the last attempt EOW? */ /* * Set at the end of recovery: the start point of a partial record at the diff --git a/src/test/recovery/t/011_crash_recovery.pl b/src/test/recovery/t/011_crash_recovery.pl index 1b57d01046d..bde16b7cfa7 100644 --- a/src/test/recovery/t/011_crash_recovery.pl +++ b/src/test/recovery/t/011_crash_recovery.pl @@ -9,7 +9,9 @@ use warnings; use PostgreSQL::Test::Cluster; use PostgreSQL::Test::Utils; use Test::More; +use IPC::Run; +my $reached_eow_pat = "reached end of WAL at "; my $node = PostgreSQL::Test::Cluster->new('primary'); $node->init(allows_streaming => 1); $node->start; @@ -47,7 +49,15 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), # Crash and restart the postmaster $node->stop('immediate'); +my $logstart = get_log_size($node); $node->start; +my $max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, $reached_eow_pat, $logstart)); + sleep 0.5; +} +ok ($max_attempts >= 0, "end-of-wal is logged"); # Make sure we really got a new xid cmp_ok($node->safe_psql('postgres', 'SELECT pg_current_xact_id()'), @@ -60,4 +70,100 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), $stdin .= "\\q\n"; $tx->finish; # wait for psql to quit gracefully +my $segsize = $node->safe_psql('postgres', + qq[SELECT setting FROM pg_settings WHERE name = 'wal_segment_size';]); + +# make sure no records afterwards go to the next segment +$node->safe_psql('postgres', qq[ + SELECT pg_switch_wal(); + CHECKPOINT; + CREATE TABLE t(); +]); +$node->stop('immediate'); + +# identify REDO WAL file +my $cmd = "pg_controldata -D " . $node->data_dir(); +$cmd = ['pg_controldata', '-D', $node->data_dir()]; +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stdout =~ /^Latest checkpoint's REDO WAL file:[ \t] *(.+)$/m, + "checkpoint file is identified"); +my $chkptfile = $1; + +# identify the last record +my $walfile = $node->data_dir() . "/pg_wal/$chkptfile"; +$cmd = ['pg_waldump', $walfile]; +$stdout = ''; +$stderr = ''; +my $lastrec; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +foreach my $l (split(/\r?\n/, $stdout)) +{ + $lastrec = $l; +} +ok(defined $lastrec, "last WAL record is extracted"); +ok($stderr =~ /end of WAL at ([0-9A-F\/]+): .* at \g1/, + "pg_waldump emits the correct ending message"); + +# read the last record LSN excluding leading zeroes +ok ($lastrec =~ /, lsn: 0\/0*([1-9A-F][0-9A-F]+),/, + "LSN of the last record identified"); +my $lastlsn = $1; + +# corrupt the last record +my $offset = hex($lastlsn) % $segsize; +open(my $segf, '+<', $walfile) or die "failed to open $walfile\n"; +seek($segf, $offset, 0); # halfway corrupt the last record +print $segf "\0\0\0\0"; +close($segf); + +# pg_waldump complains about the corrupted record +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stderr =~ /error: error in WAL record at 0\/$lastlsn: .* at 0\/$lastlsn/, + "pg_waldump emits the correct error message"); + +# also server complains +$logstart = get_log_size($node); +$node->start; +$max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, "WARNING: invalid record length at 0/$lastlsn: wanted [0-9]+, got 0", + $logstart)); + sleep 0.5; +} +ok($max_attempts >= 0, "header error is logged at $lastlsn"); + +# no end-of-wal message should be seen this time +ok(!find_in_log($node, $reached_eow_pat, $logstart), + "false log message is not emitted"); + +$node->stop('immediate'); + done_testing(); + +#### helper routines +# return the size of logfile of $node in bytes +sub get_log_size +{ + my ($node) = @_; + + return (stat $node->logfile)[7]; +} + +# find $pat in logfile of $node after $off-th byte +sub find_in_log +{ + my ($node, $pat, $off) = @_; + + $off = 0 unless defined $off; + my $log = PostgreSQL::Test::Utils::slurp_file($node->logfile); + return 0 if (length($log) <= $off); + + $log = substr($log, $off); + + return $log =~ m/$pat/; +} -- 2.25.1 --aYDVKSzuImP48n7V-- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v22] Make End-Of-Recovery error less scary @ 2022-11-15 04:41 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2022-11-15 04:41 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlogreader.c | 137 +++++++++++++++++----- src/backend/access/transam/xlogrecovery.c | 94 +++++++++++---- src/backend/replication/walreceiver.c | 7 +- src/bin/pg_waldump/pg_waldump.c | 13 +- src/include/access/xlogreader.h | 1 + src/test/recovery/t/011_crash_recovery.pl | 106 +++++++++++++++++ 6 files changed, 299 insertions(+), 59 deletions(-) diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 93f667b254..f891a62944 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -48,6 +48,8 @@ static int ReadPageInternal(XLogReaderState *state, XLogRecPtr pageptr, int reqLen); static void XLogReaderInvalReadState(XLogReaderState *state); static XLogPageReadResult XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking); +static bool ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record); static bool ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess); static bool ValidXLogRecord(XLogReaderState *state, XLogRecord *record, @@ -149,6 +151,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -558,6 +561,7 @@ XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking) /* reset error state */ state->errormsg_buf[0] = '\0'; decoded = NULL; + state->EndOfWAL = false; state->abortedRecPtr = InvalidXLogRecPtr; state->missingContrecPtr = InvalidXLogRecPtr; @@ -640,25 +644,21 @@ restart: Assert(pageHeaderSize <= readOff); /* - * Read the record length. + * Validate the record header. * - * NB: Even though we use an XLogRecord pointer here, the whole record - * header might not fit on this page. xl_tot_len is the first field of the - * struct, so it must be on this page (the records are MAXALIGNed), but we - * cannot access any other fields until we've verified that we got the - * whole header. - */ - record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); - total_len = record->xl_tot_len; - - /* - * If the whole record header is on this page, validate it immediately. - * Otherwise do just a basic sanity check on xl_tot_len, and validate the - * rest of the header after reading it from the next page. The xl_tot_len + * Even though we use an XLogRecord pointer here, the whole record header + * might not fit on this page. If the whole record header is on this page, + * validate it immediately. Even otherwise xl_tot_len must be on this page + * (it is the first field of MAXALIGNed records), but we still cannot + * access any further fields until we've verified that we got the whole + * header, so do just a basic sanity check on record length, and validate + * the rest of the header after reading it from the next page. The length * check is necessary here to ensure that we enter the "Need to reassemble * record" code path below; otherwise we might fail to apply * ValidXLogRecordHeader at all. */ + record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); + if (targetRecOff <= XLOG_BLCKSZ - SizeOfXLogRecord) { if (!ValidXLogRecordHeader(state, RecPtr, state->DecodeRecPtr, record, @@ -668,18 +668,14 @@ restart: } else { - /* XXX: more validation should be done here */ - if (total_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: wanted %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, total_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) goto err; - } + gotheader = false; } + total_len = record->xl_tot_len; + /* * Find space to decode this record. Don't allow oversized allocation if * the caller requested nonblocking. Otherwise, we *have* to try to @@ -1105,6 +1101,60 @@ XLogReaderInvalReadState(XLogReaderState *state) state->readLen = 0; } +/* + * Validate record length of an XLOG record header. + * + * This is substantially a part of ValidXLogRecordHeader. But XLogReadRecord + * needs this separate from the function in case of a partial record header. + */ +static bool +ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record) +{ + if (record->xl_tot_len == 0) + { + char *p; + char *pe; + + /* + * We are almost sure reaching the end of WAL, make sure that the + * whole page after the record is filled with zeroes. + */ + p = (char *) record; + pe = p + XLOG_BLCKSZ - (RecPtr & (XLOG_BLCKSZ - 1)); + + while (*p == 0 && p < pe) + p++; + + if (p == pe) + { + /* + * The page after the record is completely zeroed. That suggests + * we don't have a record after this point. We don't bother + * checking the pages after since they are not zeroed in the case + * of recycled segments. + */ + report_invalid_record(state, "empty record at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + return false; + } + } + + if (record->xl_tot_len < SizeOfXLogRecord) + { + report_invalid_record(state, + "invalid record length at %X/%X: wanted %u, got %u", + LSN_FORMAT_ARGS(RecPtr), + (uint32) SizeOfXLogRecord, record->xl_tot_len); + return false; + } + + return true; +} + /* * Validate an XLOG record header. * @@ -1116,14 +1166,9 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { - if (record->xl_tot_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: wanted %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, record->xl_tot_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) return false; - } + if (!RmgrIdIsValid(record->xl_rmid)) { report_invalid_record(state, @@ -1219,6 +1264,32 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, XLByteToSeg(recptr, segno, state->segcxt.ws_segsize); offset = XLogSegmentOffset(recptr, state->segcxt.ws_segsize); + StaticAssertStmt(XLOG_PAGE_MAGIC != 0, "XLOG_PAGE_MAGIC is zero"); + + if (hdr->xlp_magic == 0) + { + /* Regard an empty page as End-Of-WAL */ + int i; + + for (i = 0; i < XLOG_BLCKSZ && phdr[i] == 0; i++); + if (i == XLOG_BLCKSZ) + { + char fname[MAXFNAMELEN]; + + XLogFileName(fname, state->seg.ws_tli, segno, + state->segcxt.ws_segsize); + + report_invalid_record(state, + "empty page in log segment %s, offset %u", + fname, + offset); + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid magic number */ + } + if (hdr->xlp_magic != XLOG_PAGE_MAGIC) { char fname[MAXFNAMELEN]; @@ -1304,6 +1375,14 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, LSN_FORMAT_ARGS(hdr->xlp_pageaddr), fname, offset); + + /* + * If the page address is less than expected we assume it is an unused + * page in a recycled segment. + */ + if (hdr->xlp_pageaddr < recptr) + state->EndOfWAL = true; + return false; } diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index cb07694aea..3f54c875e5 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -1626,7 +1626,7 @@ PerformWalRecovery(void) /* just have to read next record after CheckPoint */ Assert(xlogreader->ReadRecPtr == CheckPointLoc); replayTLI = CheckPointTLI; - record = ReadRecord(xlogprefetcher, LOG, false, replayTLI); + record = ReadRecord(xlogprefetcher, WARNING, false, replayTLI); } if (record != NULL) @@ -1735,7 +1735,7 @@ PerformWalRecovery(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogprefetcher, LOG, false, replayTLI); + record = ReadRecord(xlogprefetcher, WARNING, false, replayTLI); } while (record != NULL); /* @@ -1789,11 +1789,18 @@ PerformWalRecovery(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, - (errmsg("redo is not required"))); + errmsg("redo is not required")); + } + else + { + /* broken record found */ + ereport(WARNING, + errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database.")); } /* @@ -3024,6 +3031,7 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogPrefetcherReadRecord(xlogprefetcher, &errormsg); if (record == NULL) @@ -3045,6 +3053,18 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * EndRecPtr is the LSN we tried to read but failed. In the + * case of decoding error, it is at the end of the failed + * record but we don't have a means for now to know EndRecPtr + * is pointing to which of the beginning or ending of the + * failed record. + */ + ErrRecPtr = xlogreader->EndRecPtr; } if (readFile >= 0) @@ -3055,13 +3075,15 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, /* * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * failed- in that case we already logged something. In StandbyMode + * that only happens if we have been triggered, so we shouldn't + * loop anymore in that case. When EndOfWAL is true, we don't emit + * the message immediately and instead will show it as a part of a + * decent end-of-wal message later. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), - (errmsg_internal("%s", errormsg) /* already translated */ )); + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), + errmsg_internal("%s", errormsg) /* already translated */ ); } /* @@ -3091,11 +3113,14 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, /* Great, got a record */ return record; } - else + + Assert(ErrRecPtr != InvalidXLogRecPtr); + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -3108,11 +3133,16 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + errmsg_internal("reached end of WAL at %X/%X on timeline %u in pg_wal during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI)); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -3133,12 +3163,24 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * recovery ended. + * + * Emit a decent message if we met end-of-WAL. Otherwise we should + * have already emitted an error message. + */ + if (xlogreader->EndOfWAL) + ereport(LOG, + errmsg("reached end of WAL at %X/%X on timeline %u", + LSN_FORMAT_ARGS(ErrRecPtr), replayTLI), + (errormsg ? errdetail_internal("%s", errormsg) : 0)); } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -3233,11 +3275,16 @@ retry: case XLREAD_WOULDBLOCK: return XLREAD_WOULDBLOCK; case XLREAD_FAIL: + Assert(!StandbyMode || CheckForStandbyTrigger()); + if (readFile >= 0) close(readFile); readFile = -1; readLen = 0; readSource = XLOG_FROM_ANY; + + /* promotion exit is not end-of-WAL */ + xlogreader->EndOfWAL = !StandbyMode; return XLREAD_FAIL; case XLREAD_SUCCESS: break; @@ -3898,7 +3945,8 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + /* use currentSource as readSource is reset at failure */ + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index ad383dbcaa..054a4cb127 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -498,10 +498,9 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", - startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); + errmsg("replication terminated by primary server at %X/%X on timeline %u.", + LSN_FORMAT_ARGS(LogstreamResult.Write), + startpointTLI)); endofwal = true; break; } diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index 9993378ca5..26a4125b30 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -1168,9 +1168,16 @@ main(int argc, char **argv) exit(0); if (errormsg) - pg_fatal("error in WAL record at %X/%X: %s", - LSN_FORMAT_ARGS(xlogreader_state->ReadRecPtr), - errormsg); + { + if (xlogreader_state->EndOfWAL) + pg_log_info("end of WAL at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + else + pg_fatal("error in WAL record at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + } XLogReaderFree(xlogreader_state); diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index e87f91316a..70d3b25eda 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -205,6 +205,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* was the last attempt EOW? */ /* * Set at the end of recovery: the start point of a partial record at the diff --git a/src/test/recovery/t/011_crash_recovery.pl b/src/test/recovery/t/011_crash_recovery.pl index 1b57d01046..bde16b7cfa 100644 --- a/src/test/recovery/t/011_crash_recovery.pl +++ b/src/test/recovery/t/011_crash_recovery.pl @@ -9,7 +9,9 @@ use warnings; use PostgreSQL::Test::Cluster; use PostgreSQL::Test::Utils; use Test::More; +use IPC::Run; +my $reached_eow_pat = "reached end of WAL at "; my $node = PostgreSQL::Test::Cluster->new('primary'); $node->init(allows_streaming => 1); $node->start; @@ -47,7 +49,15 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), # Crash and restart the postmaster $node->stop('immediate'); +my $logstart = get_log_size($node); $node->start; +my $max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, $reached_eow_pat, $logstart)); + sleep 0.5; +} +ok ($max_attempts >= 0, "end-of-wal is logged"); # Make sure we really got a new xid cmp_ok($node->safe_psql('postgres', 'SELECT pg_current_xact_id()'), @@ -60,4 +70,100 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), $stdin .= "\\q\n"; $tx->finish; # wait for psql to quit gracefully +my $segsize = $node->safe_psql('postgres', + qq[SELECT setting FROM pg_settings WHERE name = 'wal_segment_size';]); + +# make sure no records afterwards go to the next segment +$node->safe_psql('postgres', qq[ + SELECT pg_switch_wal(); + CHECKPOINT; + CREATE TABLE t(); +]); +$node->stop('immediate'); + +# identify REDO WAL file +my $cmd = "pg_controldata -D " . $node->data_dir(); +$cmd = ['pg_controldata', '-D', $node->data_dir()]; +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stdout =~ /^Latest checkpoint's REDO WAL file:[ \t] *(.+)$/m, + "checkpoint file is identified"); +my $chkptfile = $1; + +# identify the last record +my $walfile = $node->data_dir() . "/pg_wal/$chkptfile"; +$cmd = ['pg_waldump', $walfile]; +$stdout = ''; +$stderr = ''; +my $lastrec; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +foreach my $l (split(/\r?\n/, $stdout)) +{ + $lastrec = $l; +} +ok(defined $lastrec, "last WAL record is extracted"); +ok($stderr =~ /end of WAL at ([0-9A-F\/]+): .* at \g1/, + "pg_waldump emits the correct ending message"); + +# read the last record LSN excluding leading zeroes +ok ($lastrec =~ /, lsn: 0\/0*([1-9A-F][0-9A-F]+),/, + "LSN of the last record identified"); +my $lastlsn = $1; + +# corrupt the last record +my $offset = hex($lastlsn) % $segsize; +open(my $segf, '+<', $walfile) or die "failed to open $walfile\n"; +seek($segf, $offset, 0); # halfway corrupt the last record +print $segf "\0\0\0\0"; +close($segf); + +# pg_waldump complains about the corrupted record +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stderr =~ /error: error in WAL record at 0\/$lastlsn: .* at 0\/$lastlsn/, + "pg_waldump emits the correct error message"); + +# also server complains +$logstart = get_log_size($node); +$node->start; +$max_attempts = 360; +while ($max_attempts-- >= 0) +{ + last if (find_in_log($node, "WARNING: invalid record length at 0/$lastlsn: wanted [0-9]+, got 0", + $logstart)); + sleep 0.5; +} +ok($max_attempts >= 0, "header error is logged at $lastlsn"); + +# no end-of-wal message should be seen this time +ok(!find_in_log($node, $reached_eow_pat, $logstart), + "false log message is not emitted"); + +$node->stop('immediate'); + done_testing(); + +#### helper routines +# return the size of logfile of $node in bytes +sub get_log_size +{ + my ($node) = @_; + + return (stat $node->logfile)[7]; +} + +# find $pat in logfile of $node after $off-th byte +sub find_in_log +{ + my ($node, $pat, $off) = @_; + + $off = 0 unless defined $off; + my $log = PostgreSQL::Test::Utils::slurp_file($node->logfile); + return 0 if (length($log) <= $off); + + $log = substr($log, $off); + + return $log =~ m/$pat/; +} -- 2.31.1 ----Next_Part(Fri_Nov_18_17_25_37_2022_798)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v23] Make End-Of-Recovery error less scary @ 2022-11-30 02:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2022-11-30 02:51 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. --- src/backend/access/transam/xlogreader.c | 137 +++++++++++++++++----- src/backend/access/transam/xlogrecovery.c | 94 +++++++++++---- src/backend/replication/walreceiver.c | 7 +- src/bin/pg_waldump/pg_waldump.c | 13 +- src/include/access/xlogreader.h | 1 + src/test/recovery/t/011_crash_recovery.pl | 96 +++++++++++++++ 6 files changed, 289 insertions(+), 59 deletions(-) diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index 93f667b254..137de96795 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -48,6 +48,8 @@ static int ReadPageInternal(XLogReaderState *state, XLogRecPtr pageptr, int reqLen); static void XLogReaderInvalReadState(XLogReaderState *state); static XLogPageReadResult XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking); +static bool ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record); static bool ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess); static bool ValidXLogRecord(XLogReaderState *state, XLogRecord *record, @@ -149,6 +151,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -558,6 +561,7 @@ XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking) /* reset error state */ state->errormsg_buf[0] = '\0'; decoded = NULL; + state->EndOfWAL = false; state->abortedRecPtr = InvalidXLogRecPtr; state->missingContrecPtr = InvalidXLogRecPtr; @@ -640,25 +644,21 @@ restart: Assert(pageHeaderSize <= readOff); /* - * Read the record length. + * Validate the record header. * - * NB: Even though we use an XLogRecord pointer here, the whole record - * header might not fit on this page. xl_tot_len is the first field of the - * struct, so it must be on this page (the records are MAXALIGNed), but we - * cannot access any other fields until we've verified that we got the - * whole header. - */ - record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); - total_len = record->xl_tot_len; - - /* - * If the whole record header is on this page, validate it immediately. - * Otherwise do just a basic sanity check on xl_tot_len, and validate the - * rest of the header after reading it from the next page. The xl_tot_len + * Even though we use an XLogRecord pointer here, the whole record header + * might not fit on this page. If the whole record header is on this page, + * validate it immediately. Even otherwise xl_tot_len must be on this page + * (it is the first field of MAXALIGNed records), but we still cannot + * access any further fields until we've verified that we got the whole + * header, so do just a basic sanity check on record length, and validate + * the rest of the header after reading it from the next page. The length * check is necessary here to ensure that we enter the "Need to reassemble * record" code path below; otherwise we might fail to apply * ValidXLogRecordHeader at all. */ + record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); + if (targetRecOff <= XLOG_BLCKSZ - SizeOfXLogRecord) { if (!ValidXLogRecordHeader(state, RecPtr, state->DecodeRecPtr, record, @@ -668,18 +668,14 @@ restart: } else { - /* XXX: more validation should be done here */ - if (total_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: wanted %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, total_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) goto err; - } + gotheader = false; } + total_len = record->xl_tot_len; + /* * Find space to decode this record. Don't allow oversized allocation if * the caller requested nonblocking. Otherwise, we *have* to try to @@ -1105,6 +1101,60 @@ XLogReaderInvalReadState(XLogReaderState *state) state->readLen = 0; } +/* + * Validate record length of an XLOG record header. + * + * This is substantially a part of ValidXLogRecordHeader. But XLogReadRecord + * needs this separate from the function in case of a partial record header. + */ +static bool +ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record) +{ + if (record->xl_tot_len == 0) + { + char *p; + char *pe; + + /* + * We are almost sure reaching the end of WAL, make sure that the + * whole page after the record is filled with zeroes. + */ + p = (char *) record; + pe = p + XLOG_BLCKSZ - (RecPtr & (XLOG_BLCKSZ - 1)); + + while (p < pe && *p == 0) + p++; + + if (p == pe) + { + /* + * The page after the record is completely zeroed. That suggests + * we don't have a record after this point. We don't bother + * checking the pages after since they are not zeroed in the case + * of recycled segments. + */ + report_invalid_record(state, "empty record at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + return false; + } + } + + if (record->xl_tot_len < SizeOfXLogRecord) + { + report_invalid_record(state, + "invalid record length at %X/%X: wanted %u, got %u", + LSN_FORMAT_ARGS(RecPtr), + (uint32) SizeOfXLogRecord, record->xl_tot_len); + return false; + } + + return true; +} + /* * Validate an XLOG record header. * @@ -1116,14 +1166,9 @@ ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { - if (record->xl_tot_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: wanted %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, record->xl_tot_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) return false; - } + if (!RmgrIdIsValid(record->xl_rmid)) { report_invalid_record(state, @@ -1219,6 +1264,32 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, XLByteToSeg(recptr, segno, state->segcxt.ws_segsize); offset = XLogSegmentOffset(recptr, state->segcxt.ws_segsize); + StaticAssertStmt(XLOG_PAGE_MAGIC != 0, "XLOG_PAGE_MAGIC is zero"); + + if (hdr->xlp_magic == 0) + { + /* Regard an empty page as End-Of-WAL */ + int i; + + for (i = 0; i < XLOG_BLCKSZ && phdr[i] == 0; i++); + if (i == XLOG_BLCKSZ) + { + char fname[MAXFNAMELEN]; + + XLogFileName(fname, state->seg.ws_tli, segno, + state->segcxt.ws_segsize); + + report_invalid_record(state, + "empty page in log segment %s, offset %u", + fname, + offset); + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid magic number */ + } + if (hdr->xlp_magic != XLOG_PAGE_MAGIC) { char fname[MAXFNAMELEN]; @@ -1304,6 +1375,14 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, LSN_FORMAT_ARGS(hdr->xlp_pageaddr), fname, offset); + + /* + * If the page address is less than expected we assume it is an unused + * page in a recycled segment. + */ + if (hdr->xlp_pageaddr < recptr) + state->EndOfWAL = true; + return false; } diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index 41ffc57da9..049b468620 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -1625,7 +1625,7 @@ PerformWalRecovery(void) /* just have to read next record after CheckPoint */ Assert(xlogreader->ReadRecPtr == CheckPointLoc); replayTLI = CheckPointTLI; - record = ReadRecord(xlogprefetcher, LOG, false, replayTLI); + record = ReadRecord(xlogprefetcher, WARNING, false, replayTLI); } if (record != NULL) @@ -1734,7 +1734,7 @@ PerformWalRecovery(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogprefetcher, LOG, false, replayTLI); + record = ReadRecord(xlogprefetcher, WARNING, false, replayTLI); } while (record != NULL); /* @@ -1788,11 +1788,18 @@ PerformWalRecovery(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, - (errmsg("redo is not required"))); + errmsg("redo is not required")); + } + else + { + /* broken record found */ + ereport(WARNING, + errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database.")); } /* @@ -3020,6 +3027,7 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogPrefetcherReadRecord(xlogprefetcher, &errormsg); if (record == NULL) @@ -3041,6 +3049,18 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * EndRecPtr is the LSN we tried to read but failed. In the + * case of decoding error, it is at the end of the failed + * record but we don't have a means for now to know EndRecPtr + * is pointing to which of the beginning or ending of the + * failed record. + */ + ErrRecPtr = xlogreader->EndRecPtr; } if (readFile >= 0) @@ -3051,13 +3071,15 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, /* * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * failed- in that case we already logged something. In StandbyMode + * that only happens if we have been triggered, so we shouldn't + * loop anymore in that case. When EndOfWAL is true, we don't emit + * the message immediately and instead will show it as a part of a + * decent end-of-wal message later. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), - (errmsg_internal("%s", errormsg) /* already translated */ )); + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), + errmsg_internal("%s", errormsg) /* already translated */ ); } /* @@ -3087,11 +3109,14 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, /* Great, got a record */ return record; } - else + + Assert(ErrRecPtr != InvalidXLogRecPtr); + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -3104,11 +3129,16 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + errmsg_internal("reached end of WAL at %X/%X on timeline %u in pg_wal during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI)); InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -3129,12 +3159,24 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * recovery ended. + * + * Emit a decent message if we met end-of-WAL. Otherwise we should + * have already emitted an error message. + */ + if (xlogreader->EndOfWAL) + ereport(LOG, + errmsg("reached end of WAL at %X/%X on timeline %u", + LSN_FORMAT_ARGS(ErrRecPtr), replayTLI), + (errormsg ? errdetail_internal("%s", errormsg) : 0)); } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -3229,11 +3271,16 @@ retry: case XLREAD_WOULDBLOCK: return XLREAD_WOULDBLOCK; case XLREAD_FAIL: + Assert(!StandbyMode || CheckForStandbyTrigger()); + if (readFile >= 0) close(readFile); readFile = -1; readLen = 0; readSource = XLOG_FROM_ANY; + + /* promotion exit is not end-of-WAL */ + xlogreader->EndOfWAL = !StandbyMode; return XLREAD_FAIL; case XLREAD_SUCCESS: break; @@ -3899,7 +3946,8 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + /* use currentSource as readSource is reset at failure */ + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index ad383dbcaa..054a4cb127 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -498,10 +498,9 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", - startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); + errmsg("replication terminated by primary server at %X/%X on timeline %u.", + LSN_FORMAT_ARGS(LogstreamResult.Write), + startpointTLI)); endofwal = true; break; } diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index 9993378ca5..26a4125b30 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -1168,9 +1168,16 @@ main(int argc, char **argv) exit(0); if (errormsg) - pg_fatal("error in WAL record at %X/%X: %s", - LSN_FORMAT_ARGS(xlogreader_state->ReadRecPtr), - errormsg); + { + if (xlogreader_state->EndOfWAL) + pg_log_info("end of WAL at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + else + pg_fatal("error in WAL record at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + } XLogReaderFree(xlogreader_state); diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index e87f91316a..70d3b25eda 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -205,6 +205,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* was the last attempt EOW? */ /* * Set at the end of recovery: the start point of a partial record at the diff --git a/src/test/recovery/t/011_crash_recovery.pl b/src/test/recovery/t/011_crash_recovery.pl index 1b57d01046..8a379e8988 100644 --- a/src/test/recovery/t/011_crash_recovery.pl +++ b/src/test/recovery/t/011_crash_recovery.pl @@ -9,7 +9,9 @@ use warnings; use PostgreSQL::Test::Cluster; use PostgreSQL::Test::Utils; use Test::More; +use IPC::Run; +my $reached_eow_pat = "reached end of WAL at "; my $node = PostgreSQL::Test::Cluster->new('primary'); $node->init(allows_streaming => 1); $node->start; @@ -47,7 +49,10 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), # Crash and restart the postmaster $node->stop('immediate'); +my $logstart = get_log_size($node); $node->start; +$node->wait_for_log($reached_eow_pat, $logstart); +pass("end-of-wal is logged"); # Make sure we really got a new xid cmp_ok($node->safe_psql('postgres', 'SELECT pg_current_xact_id()'), @@ -60,4 +65,95 @@ is($node->safe_psql('postgres', qq[SELECT pg_xact_status('$xid');]), $stdin .= "\\q\n"; $tx->finish; # wait for psql to quit gracefully +my $segsize = $node->safe_psql('postgres', + qq[SELECT setting FROM pg_settings WHERE name = 'wal_segment_size';]); + +# make sure no records afterwards go to the next segment +$node->safe_psql('postgres', qq[ + SELECT pg_switch_wal(); + CHECKPOINT; + CREATE TABLE t(); +]); +$node->stop('immediate'); + +# identify REDO WAL file +my $cmd = "pg_controldata -D " . $node->data_dir(); +$cmd = ['pg_controldata', '-D', $node->data_dir()]; +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stdout =~ /^Latest checkpoint's REDO WAL file:[ \t] *(.+)$/m, + "checkpoint file is identified"); +my $chkptfile = $1; + +# identify the last record +my $walfile = $node->data_dir() . "/pg_wal/$chkptfile"; +$cmd = ['pg_waldump', $walfile]; +$stdout = ''; +$stderr = ''; +my $lastrec; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +foreach my $l (split(/\r?\n/, $stdout)) +{ + $lastrec = $l; +} +ok(defined $lastrec, "last WAL record is extracted"); +ok($stderr =~ /end of WAL at ([0-9A-F\/]+): .* at \g1/, + "pg_waldump emits the correct ending message"); + +# read the last record LSN excluding leading zeroes +ok ($lastrec =~ /, lsn: 0\/0*([1-9A-F][0-9A-F]+),/, + "LSN of the last record identified"); +my $lastlsn = $1; + +# corrupt the last record +my $offset = hex($lastlsn) % $segsize; +open(my $segf, '+<', $walfile) or die "failed to open $walfile\n"; +seek($segf, $offset, 0); # halfway corrupt the last record +print $segf "\0\0\0\0"; +close($segf); + +# pg_waldump complains about the corrupted record +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stderr =~ /error: error in WAL record at 0\/$lastlsn: .* at 0\/$lastlsn/, + "pg_waldump emits the correct error message"); + +# also server complains +$logstart = get_log_size($node); +$node->start; +$node->wait_for_log("WARNING: invalid record length at 0/$lastlsn: wanted [0-9]+, got 0", + $logstart); +pass("header error is logged at $lastlsn"); + +# no end-of-wal message should be seen this time +ok(!find_in_log($node, $reached_eow_pat, $logstart), + "false log message is not emitted"); + +$node->stop('immediate'); + done_testing(); + +#### helper routines +# return the size of logfile of $node in bytes +sub get_log_size +{ + my ($node) = @_; + + return (stat $node->logfile)[7]; +} + +# find $pat in logfile of $node after $off-th byte +sub find_in_log +{ + my ($node, $pat, $off) = @_; + + $off = 0 unless defined $off; + my $log = PostgreSQL::Test::Utils::slurp_file($node->logfile); + return 0 if (length($log) <= $off); + + $log = substr($log, $off); + + return $log =~ m/$pat/; +} -- 2.31.1 ----Next_Part(Wed_Nov_30_11_56_02_2022_262)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v24] Make End-Of-Recovery error less scary @ 2022-11-30 02:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2022-11-30 02:51 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. To make sure that the detection is correct, this patch checks if all trailing bytes in the same page are zeroed in that case. --- src/backend/access/transam/xlogreader.c | 144 +++++++++++++++++----- src/backend/access/transam/xlogrecovery.c | 94 ++++++++++---- src/backend/replication/walreceiver.c | 7 +- src/bin/pg_waldump/pg_waldump.c | 13 +- src/include/access/xlogreader.h | 1 + src/test/recovery/t/034_recovery.pl | 135 ++++++++++++++++++++ 6 files changed, 335 insertions(+), 59 deletions(-) create mode 100644 src/test/recovery/t/034_recovery.pl diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index aa6c929477..8cb2d55333 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -48,6 +48,8 @@ static int ReadPageInternal(XLogReaderState *state, XLogRecPtr pageptr, int reqLen); static void XLogReaderInvalReadState(XLogReaderState *state); static XLogPageReadResult XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking); +static bool ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record); static bool ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess); static bool ValidXLogRecord(XLogReaderState *state, XLogRecord *record, @@ -149,6 +151,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -558,6 +561,7 @@ XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking) /* reset error state */ state->errormsg_buf[0] = '\0'; decoded = NULL; + state->EndOfWAL = false; state->abortedRecPtr = InvalidXLogRecPtr; state->missingContrecPtr = InvalidXLogRecPtr; @@ -640,25 +644,21 @@ restart: Assert(pageHeaderSize <= readOff); /* - * Read the record length. + * Validate the record header. * - * NB: Even though we use an XLogRecord pointer here, the whole record - * header might not fit on this page. xl_tot_len is the first field of the - * struct, so it must be on this page (the records are MAXALIGNed), but we - * cannot access any other fields until we've verified that we got the - * whole header. - */ - record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); - total_len = record->xl_tot_len; - - /* - * If the whole record header is on this page, validate it immediately. - * Otherwise do just a basic sanity check on xl_tot_len, and validate the - * rest of the header after reading it from the next page. The xl_tot_len + * Even though we use an XLogRecord pointer here, the whole record header + * might not fit on this page. If the whole record header is on this page, + * validate it immediately. Even otherwise xl_tot_len must be on this page + * (it is the first field of MAXALIGNed records), but we still cannot + * access any further fields until we've verified that we got the whole + * header, so do just a basic sanity check on record length, and validate + * the rest of the header after reading it from the next page. The length * check is necessary here to ensure that we enter the "Need to reassemble * record" code path below; otherwise we might fail to apply * ValidXLogRecordHeader at all. */ + record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); + if (targetRecOff <= XLOG_BLCKSZ - SizeOfXLogRecord) { if (!ValidXLogRecordHeader(state, RecPtr, state->DecodeRecPtr, record, @@ -668,18 +668,14 @@ restart: } else { - /* XXX: more validation should be done here */ - if (total_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: wanted %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, total_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) goto err; - } + gotheader = false; } + total_len = record->xl_tot_len; + /* * Find space to decode this record. Don't allow oversized allocation if * the caller requested nonblocking. Otherwise, we *have* to try to @@ -1105,25 +1101,81 @@ XLogReaderInvalReadState(XLogReaderState *state) state->readLen = 0; } +/* + * Validate record length of an XLOG record header. + * + * This is substantially a part of ValidXLogRecordHeader. But XLogReadRecord + * needs this separate from the function in case of a partial record header. + * + * Returns true if the xl_tot_len header field has a seemingly valid value, + * which means the caller can proceed reading to the following part of the + * record. + */ +static bool +ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record) +{ + if (record->xl_tot_len == 0) + { + char *p; + char *pe; + + /* + * We are almost sure reaching the end of WAL, make sure that the + * whole page after the record is filled with zeroes. + */ + p = (char *) record; + pe = p + XLOG_BLCKSZ - (RecPtr & (XLOG_BLCKSZ - 1)); + + while (p < pe && *p == 0) + p++; + + if (p == pe) + { + /* + * The page after the record is completely zeroed. That suggests + * we don't have a record after this point. We don't bother + * checking the pages after since they are not zeroed in the case + * of recycled segments. + */ + report_invalid_record(state, "empty record at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + return false; + } + } + + if (record->xl_tot_len < SizeOfXLogRecord) + { + report_invalid_record(state, + "invalid record length at %X/%X: wanted %u, got %u", + LSN_FORMAT_ARGS(RecPtr), + (uint32) SizeOfXLogRecord, record->xl_tot_len); + return false; + } + + return true; +} + /* * Validate an XLOG record header. * * This is just a convenience subroutine to avoid duplicated code in * XLogReadRecord. It's not intended for use from anywhere else. + * + * Returns true if the header fields have the valid values and the caller can + * proceed reading to the following part of the record. */ static bool ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { - if (record->xl_tot_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: wanted %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, record->xl_tot_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) return false; - } + if (!RmgrIdIsValid(record->xl_rmid)) { report_invalid_record(state, @@ -1219,6 +1271,32 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, XLByteToSeg(recptr, segno, state->segcxt.ws_segsize); offset = XLogSegmentOffset(recptr, state->segcxt.ws_segsize); + StaticAssertStmt(XLOG_PAGE_MAGIC != 0, "XLOG_PAGE_MAGIC is zero"); + + if (hdr->xlp_magic == 0) + { + /* Regard an empty page as End-Of-WAL */ + int i; + + for (i = 0; i < XLOG_BLCKSZ && phdr[i] == 0; i++); + if (i == XLOG_BLCKSZ) + { + char fname[MAXFNAMELEN]; + + XLogFileName(fname, state->seg.ws_tli, segno, + state->segcxt.ws_segsize); + + report_invalid_record(state, + "empty page in log segment %s, offset %u", + fname, + offset); + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid magic number */ + } + if (hdr->xlp_magic != XLOG_PAGE_MAGIC) { char fname[MAXFNAMELEN]; @@ -1308,6 +1386,14 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, fname, LSN_FORMAT_ARGS(recptr), offset); + + /* + * If the page address is less than expected we assume it is an unused + * page in a recycled segment. + */ + if (hdr->xlp_pageaddr < recptr) + state->EndOfWAL = true; + return false; } diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index dbe9394762..3ada8b346d 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -1644,7 +1644,7 @@ PerformWalRecovery(void) /* just have to read next record after CheckPoint */ Assert(xlogreader->ReadRecPtr == CheckPointLoc); replayTLI = CheckPointTLI; - record = ReadRecord(xlogprefetcher, LOG, false, replayTLI); + record = ReadRecord(xlogprefetcher, WARNING, false, replayTLI); } if (record != NULL) @@ -1753,7 +1753,7 @@ PerformWalRecovery(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogprefetcher, LOG, false, replayTLI); + record = ReadRecord(xlogprefetcher, WARNING, false, replayTLI); } while (record != NULL); /* @@ -1807,11 +1807,18 @@ PerformWalRecovery(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, - (errmsg("redo is not required"))); + errmsg("redo is not required")); + } + else + { + /* broken record found */ + ereport(WARNING, + errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database.")); } /* @@ -3044,6 +3051,7 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogPrefetcherReadRecord(xlogprefetcher, &errormsg); if (record == NULL) @@ -3065,6 +3073,18 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * EndRecPtr is the LSN we tried to read but failed. In the + * case of decoding error, it is at the end of the failed + * record but we don't have a means for now to know EndRecPtr + * is pointing to which of the beginning or ending of the + * failed record. + */ + ErrRecPtr = xlogreader->EndRecPtr; } if (readFile >= 0) @@ -3075,13 +3095,15 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, /* * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * failed- in that case we already logged something. In StandbyMode + * that only happens if we have been triggered, so we shouldn't + * loop anymore in that case. When EndOfWAL is true, we don't emit + * the message immediately and instead will show it as a part of a + * decent end-of-wal message later. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), - (errmsg_internal("%s", errormsg) /* already translated */ )); + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), + errmsg_internal("%s", errormsg) /* already translated */ ); } /* @@ -3112,11 +3134,14 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, /* Great, got a record */ return record; } - else + + Assert(ErrRecPtr != InvalidXLogRecPtr); + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -3129,11 +3154,16 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + errmsg_internal("reached end of WAL at %X/%X on timeline %u in pg_wal during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI)); InArchiveRecovery = true; if (StandbyModeRequested) EnableStandbyMode(); @@ -3154,12 +3184,24 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * recovery ended. + * + * Emit a decent message if we met end-of-WAL. Otherwise we should + * have already emitted an error message. + */ + if (xlogreader->EndOfWAL) + ereport(LOG, + errmsg("reached end of WAL at %X/%X on timeline %u", + LSN_FORMAT_ARGS(ErrRecPtr), replayTLI), + (errormsg ? errdetail_internal("%s", errormsg) : 0)); } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -3256,11 +3298,16 @@ retry: case XLREAD_WOULDBLOCK: return XLREAD_WOULDBLOCK; case XLREAD_FAIL: + Assert(!StandbyMode || CheckForStandbyTrigger()); + if (readFile >= 0) close(readFile); readFile = -1; readLen = 0; readSource = XLOG_FROM_ANY; + + /* promotion exit is not end-of-WAL */ + xlogreader->EndOfWAL = !StandbyMode; return XLREAD_FAIL; case XLREAD_SUCCESS: break; @@ -3928,7 +3975,8 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + /* use currentSource as readSource is reset at failure */ + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index f6446da2d6..78b85c5a25 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -498,10 +498,9 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", - startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); + errmsg("replication terminated by primary server at %X/%X on timeline %u.", + LSN_FORMAT_ARGS(LogstreamResult.Write), + startpointTLI)); endofwal = true; break; } diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index 44b5c8726e..0f0cccd425 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -1276,9 +1276,16 @@ main(int argc, char **argv) exit(0); if (errormsg) - pg_fatal("error in WAL record at %X/%X: %s", - LSN_FORMAT_ARGS(xlogreader_state->ReadRecPtr), - errormsg); + { + if (xlogreader_state->EndOfWAL) + pg_log_info("end of WAL at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + else + pg_fatal("error in WAL record at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + } XLogReaderFree(xlogreader_state); diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index d77bb2ab9b..84562d9de1 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -205,6 +205,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* was the last attempt EOW? */ /* * Set at the end of recovery: the start point of a partial record at the diff --git a/src/test/recovery/t/034_recovery.pl b/src/test/recovery/t/034_recovery.pl new file mode 100644 index 0000000000..580ae3b9f1 --- /dev/null +++ b/src/test/recovery/t/034_recovery.pl @@ -0,0 +1,135 @@ + +# Copyright (c) 2021-2023, PostgreSQL Global Development Group + +# Minimal test testing recovery process + +use strict; +use warnings; +use PostgreSQL::Test::Cluster; +use PostgreSQL::Test::Utils; +use Test::More; +use IPC::Run; + +my $reached_eow_pat = "reached end of WAL at "; +my $node = PostgreSQL::Test::Cluster->new('primary'); +$node->init(allows_streaming => 1); +$node->start; + +my ($stdout, $stderr) = ('', ''); + +my $segsize = $node->safe_psql('postgres', + qq[SELECT setting FROM pg_settings WHERE name = 'wal_segment_size';]); + +# make sure no records afterwards go to the next segment +$node->safe_psql('postgres', qq[ + SELECT pg_switch_wal(); + CHECKPOINT; + CREATE TABLE t(); +]); +$node->stop('immediate'); + +# identify REDO WAL file +my $cmd = "pg_controldata -D " . $node->data_dir(); +$cmd = ['pg_controldata', '-D', $node->data_dir()]; +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stdout =~ /^Latest checkpoint's REDO WAL file:[ \t] *(.+)$/m, + "checkpoint file is identified"); +my $chkptfile = $1; + +# identify the last record +my $walfile = $node->data_dir() . "/pg_wal/$chkptfile"; +$cmd = ['pg_waldump', $walfile]; +$stdout = ''; +$stderr = ''; +my $lastrec; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +foreach my $l (split(/\r?\n/, $stdout)) +{ + $lastrec = $l; +} +ok(defined $lastrec, "last WAL record is extracted"); +ok($stderr =~ /end of WAL at ([0-9A-F\/]+): .* at \g1/, + "pg_waldump emits the correct ending message"); + +# read the last record LSN excluding leading zeroes +ok ($lastrec =~ /, lsn: 0\/0*([1-9A-F][0-9A-F]+),/, + "LSN of the last record identified"); +my $lastlsn = $1; + +# corrupt the last record +my $offset = hex($lastlsn) % $segsize; +open(my $segf, '+<', $walfile) or die "failed to open $walfile\n"; +seek($segf, $offset, 0); # zero xl_tot_len, leaving following bytes alone. +print $segf "\0\0\0\0"; +close($segf); + +# pg_waldump complains about the corrupted record +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stderr =~ /error: error in WAL record at 0\/$lastlsn: .* at 0\/$lastlsn/, + "pg_waldump emits the correct error message"); + +# also server complains for the same reason +my $logstart = get_log_size($node); +$node->start; +$node->wait_for_log("WARNING: invalid record length at 0/$lastlsn: wanted [0-9]+, got 0", + $logstart); +pass("header error is logged at $lastlsn"); + +# no end-of-wal message should be seen this time +ok(!find_in_log($node, $reached_eow_pat, $logstart), + "false log message is not emitted"); + +# Create streaming standby linking to primary +my $backup_name = 'my_backup'; +$node->backup($backup_name); +my $node_standby = PostgreSQL::Test::Cluster->new('standby'); +$node_standby->init_from_backup($node, $backup_name, has_streaming => 1); +$node_standby->start; +$node->safe_psql('postgres', 'CREATE TABLE t ()'); +my $primary_lsn = $node->lsn('write'); +$node->wait_for_catchup($node_standby, 'write', $primary_lsn); + +$node_standby->stop(); +$node->stop('immediate'); + +# crash restart the primary +$logstart = get_log_size($node); +$node->start(); +$node->wait_for_log($reached_eow_pat, $logstart); + +# restart the standby +$logstart = get_log_size($node_standby); +$node_standby->start(); +$node_standby->wait_for_log($reached_eow_pat, $logstart); + +$node_standby->stop(); +$node->stop(); + +done_testing(); + +#### helper routines +# return the size of logfile of $node in bytes +sub get_log_size +{ + my ($node) = @_; + + return (stat $node->logfile)[7]; +} + +# find $pat in logfile of $node after $off-th byte +sub find_in_log +{ + my ($node, $pat, $off) = @_; + + $off = 0 unless defined $off; + my $log = PostgreSQL::Test::Utils::slurp_file($node->logfile); + return 0 if (length($log) <= $off); + + $log = substr($log, $off); + + return $log =~ m/$pat/; +} -- 2.31.1 ----Next_Part(Tue_Feb__7_16_07_03_2023_615)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* [PATCH v25] Make End-Of-Recovery error less scary @ 2023-03-07 05:55 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Kyotaro Horiguchi @ 2023-03-07 05:55 UTC (permalink / raw) When recovery in any type ends, we see a bit scary error message like "invalid record length" that suggests something serious is happening. Actually if recovery meets a record with length = 0, that usually means it finished applying all available WAL records. Make this message less scary as "reached end of WAL". Instead, raise the error level for other kind of WAL failure to WARNING. To make sure that the detection is correct, this patch checks if all trailing bytes in the same page are zeroed in that case. --- src/backend/access/transam/xlogreader.c | 135 ++++++++++++++++++---- src/backend/access/transam/xlogrecovery.c | 94 +++++++++++---- src/backend/replication/walreceiver.c | 7 +- src/bin/pg_waldump/pg_waldump.c | 13 ++- src/include/access/xlogreader.h | 1 + src/test/recovery/t/035_recovery.pl | 130 +++++++++++++++++++++ 6 files changed, 329 insertions(+), 51 deletions(-) create mode 100644 src/test/recovery/t/035_recovery.pl diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index cadea21b37..5a27c10bbb 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -48,6 +48,8 @@ static int ReadPageInternal(XLogReaderState *state, XLogRecPtr pageptr, int reqLen); static void XLogReaderInvalReadState(XLogReaderState *state); static XLogPageReadResult XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking); +static bool ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record); static bool ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess); static bool ValidXLogRecord(XLogReaderState *state, XLogRecord *record, @@ -149,6 +151,7 @@ XLogReaderAllocate(int wal_segment_size, const char *waldir, pfree(state); return NULL; } + state->EndOfWAL = false; state->errormsg_buf[0] = '\0'; /* @@ -558,6 +561,7 @@ XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking) /* reset error state */ state->errormsg_buf[0] = '\0'; decoded = NULL; + state->EndOfWAL = false; state->abortedRecPtr = InvalidXLogRecPtr; state->missingContrecPtr = InvalidXLogRecPtr; @@ -641,16 +645,12 @@ restart: Assert(pageHeaderSize <= readOff); /* - * Read the record length. + * Verify the record header. * * NB: Even though we use an XLogRecord pointer here, the whole record - * header might not fit on this page. xl_tot_len is the first field of the - * struct, so it must be on this page (the records are MAXALIGNed), but we - * cannot access any other fields until we've verified that we got the - * whole header. + * header might not fit on this page. */ record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ); - total_len = record->xl_tot_len; /* * If the whole record header is on this page, validate it immediately. @@ -669,18 +669,21 @@ restart: } else { - /* XXX: more validation should be done here */ - if (total_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: expected at least %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, total_len); + /* + * xl_tot_len is the first field of the struct, so it must be on this + * page (the records are MAXALIGNed), but we cannot access any other + * fields until we've verified that we got the whole header. + * + * XXX: more validation should be done here + */ + if (!ValidXLogRecordLength(state, RecPtr, record)) goto err; - } + gotheader = false; } + total_len = record->xl_tot_len; + /* * Find space to decode this record. Don't allow oversized allocation if * the caller requested nonblocking. Otherwise, we *have* to try to @@ -1106,25 +1109,81 @@ XLogReaderInvalReadState(XLogReaderState *state) state->readLen = 0; } +/* + * Validate record length of an XLOG record header. + * + * This is substantially a part of ValidXLogRecordHeader. But XLogReadRecord + * needs this separate from the function in case of a partial record header. + * + * Returns true if the xl_tot_len header field has a seemingly valid value, + * which means the caller can proceed reading to the following part of the + * record. + */ +static bool +ValidXLogRecordLength(XLogReaderState *state, XLogRecPtr RecPtr, + XLogRecord *record) +{ + if (record->xl_tot_len == 0) + { + char *p; + char *pe; + + /* + * We are almost sure reaching the end of WAL, make sure that the + * whole page after the record is filled with zeroes. + */ + p = (char *) record; + pe = p + XLOG_BLCKSZ - (RecPtr & (XLOG_BLCKSZ - 1)); + + while (p < pe && *p == 0) + p++; + + if (p == pe) + { + /* + * The page after the record is completely zeroed. That suggests + * we don't have a record after this point. We don't bother + * checking the pages after since they are not zeroed in the case + * of recycled segments. + */ + report_invalid_record(state, "empty record at %X/%X", + LSN_FORMAT_ARGS(RecPtr)); + + /* notify end-of-wal to callers */ + state->EndOfWAL = true; + return false; + } + } + + if (record->xl_tot_len < SizeOfXLogRecord) + { + report_invalid_record(state, + "invalid record length at %X/%X: expected at least %u, got %u", + LSN_FORMAT_ARGS(RecPtr), + (uint32) SizeOfXLogRecord, record->xl_tot_len); + return false; + } + + return true; +} + /* * Validate an XLOG record header. * * This is just a convenience subroutine to avoid duplicated code in * XLogReadRecord. It's not intended for use from anywhere else. + * + * Returns true if the header fields have the valid values and the caller can + * proceed reading to the following part of the record. */ static bool ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr, XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess) { - if (record->xl_tot_len < SizeOfXLogRecord) - { - report_invalid_record(state, - "invalid record length at %X/%X: expected at least %u, got %u", - LSN_FORMAT_ARGS(RecPtr), - (uint32) SizeOfXLogRecord, record->xl_tot_len); + if (!ValidXLogRecordLength(state, RecPtr, record)) return false; - } + if (!RmgrIdIsValid(record->xl_rmid)) { report_invalid_record(state, @@ -1220,6 +1279,32 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, XLByteToSeg(recptr, segno, state->segcxt.ws_segsize); offset = XLogSegmentOffset(recptr, state->segcxt.ws_segsize); + StaticAssertStmt(XLOG_PAGE_MAGIC != 0, "XLOG_PAGE_MAGIC is zero"); + + if (hdr->xlp_magic == 0) + { + /* Regard an empty page as End-Of-WAL */ + int i; + + for (i = 0; i < XLOG_BLCKSZ && phdr[i] == 0; i++); + if (i == XLOG_BLCKSZ) + { + char fname[MAXFNAMELEN]; + + XLogFileName(fname, state->seg.ws_tli, segno, + state->segcxt.ws_segsize); + + report_invalid_record(state, + "empty page in log segment %s, offset %u", + fname, + offset); + state->EndOfWAL = true; + return false; + } + + /* The same condition will be caught as invalid magic number */ + } + if (hdr->xlp_magic != XLOG_PAGE_MAGIC) { char fname[MAXFNAMELEN]; @@ -1309,6 +1394,14 @@ XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr, fname, LSN_FORMAT_ARGS(recptr), offset); + + /* + * If the page address is less than expected we assume it is an unused + * page in a recycled segment. + */ + if (hdr->xlp_pageaddr < recptr) + state->EndOfWAL = true; + return false; } diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index dbe9394762..3ada8b346d 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -1644,7 +1644,7 @@ PerformWalRecovery(void) /* just have to read next record after CheckPoint */ Assert(xlogreader->ReadRecPtr == CheckPointLoc); replayTLI = CheckPointTLI; - record = ReadRecord(xlogprefetcher, LOG, false, replayTLI); + record = ReadRecord(xlogprefetcher, WARNING, false, replayTLI); } if (record != NULL) @@ -1753,7 +1753,7 @@ PerformWalRecovery(void) } /* Else, try to fetch the next WAL record */ - record = ReadRecord(xlogprefetcher, LOG, false, replayTLI); + record = ReadRecord(xlogprefetcher, WARNING, false, replayTLI); } while (record != NULL); /* @@ -1807,11 +1807,18 @@ PerformWalRecovery(void) InRedo = false; } - else + else if (xlogreader->EndOfWAL) { /* there are no WAL records following the checkpoint */ ereport(LOG, - (errmsg("redo is not required"))); + errmsg("redo is not required")); + } + else + { + /* broken record found */ + ereport(WARNING, + errmsg("redo is skipped"), + errhint("This suggests WAL file corruption. You might need to check the database.")); } /* @@ -3044,6 +3051,7 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, for (;;) { char *errormsg; + XLogRecPtr ErrRecPtr = InvalidXLogRecPtr; record = XLogPrefetcherReadRecord(xlogprefetcher, &errormsg); if (record == NULL) @@ -3065,6 +3073,18 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, { abortedRecPtr = xlogreader->abortedRecPtr; missingContrecPtr = xlogreader->missingContrecPtr; + ErrRecPtr = abortedRecPtr; + } + else + { + /* + * EndRecPtr is the LSN we tried to read but failed. In the + * case of decoding error, it is at the end of the failed + * record but we don't have a means for now to know EndRecPtr + * is pointing to which of the beginning or ending of the + * failed record. + */ + ErrRecPtr = xlogreader->EndRecPtr; } if (readFile >= 0) @@ -3075,13 +3095,15 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, /* * We only end up here without a message when XLogPageRead() - * failed - in that case we already logged something. In - * StandbyMode that only happens if we have been triggered, so we - * shouldn't loop anymore in that case. + * failed- in that case we already logged something. In StandbyMode + * that only happens if we have been triggered, so we shouldn't + * loop anymore in that case. When EndOfWAL is true, we don't emit + * the message immediately and instead will show it as a part of a + * decent end-of-wal message later. */ - if (errormsg) - ereport(emode_for_corrupt_record(emode, xlogreader->EndRecPtr), - (errmsg_internal("%s", errormsg) /* already translated */ )); + if (!xlogreader->EndOfWAL && errormsg) + ereport(emode_for_corrupt_record(emode, ErrRecPtr), + errmsg_internal("%s", errormsg) /* already translated */ ); } /* @@ -3112,11 +3134,14 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, /* Great, got a record */ return record; } - else + + Assert(ErrRecPtr != InvalidXLogRecPtr); + + /* No valid record available from this source */ + lastSourceFailed = true; + + if (!fetching_ckpt) { - /* No valid record available from this source */ - lastSourceFailed = true; - /* * If archive recovery was requested, but we were still doing * crash recovery, switch to archive recovery and retry using the @@ -3129,11 +3154,16 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, * we'd have no idea how far we'd have to replay to reach * consistency. So err on the safe side and give up. */ - if (!InArchiveRecovery && ArchiveRecoveryRequested && - !fetching_ckpt) + if (!InArchiveRecovery && ArchiveRecoveryRequested) { + /* + * We don't report this as LOG, since we don't stop recovery + * here + */ ereport(DEBUG1, - (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + errmsg_internal("reached end of WAL at %X/%X on timeline %u in pg_wal during crash recovery, entering archive recovery", + LSN_FORMAT_ARGS(ErrRecPtr), + replayTLI)); InArchiveRecovery = true; if (StandbyModeRequested) EnableStandbyMode(); @@ -3154,12 +3184,24 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, continue; } - /* In standby mode, loop back to retry. Otherwise, give up. */ - if (StandbyMode && !CheckForStandbyTrigger()) - continue; - else - return NULL; + /* + * recovery ended. + * + * Emit a decent message if we met end-of-WAL. Otherwise we should + * have already emitted an error message. + */ + if (xlogreader->EndOfWAL) + ereport(LOG, + errmsg("reached end of WAL at %X/%X on timeline %u", + LSN_FORMAT_ARGS(ErrRecPtr), replayTLI), + (errormsg ? errdetail_internal("%s", errormsg) : 0)); } + + /* In standby mode, loop back to retry. Otherwise, give up. */ + if (StandbyMode && !CheckForStandbyTrigger()) + continue; + else + return NULL; } } @@ -3256,11 +3298,16 @@ retry: case XLREAD_WOULDBLOCK: return XLREAD_WOULDBLOCK; case XLREAD_FAIL: + Assert(!StandbyMode || CheckForStandbyTrigger()); + if (readFile >= 0) close(readFile); readFile = -1; readLen = 0; readSource = XLOG_FROM_ANY; + + /* promotion exit is not end-of-WAL */ + xlogreader->EndOfWAL = !StandbyMode; return XLREAD_FAIL; case XLREAD_SUCCESS: break; @@ -3928,7 +3975,8 @@ emode_for_corrupt_record(int emode, XLogRecPtr RecPtr) { static XLogRecPtr lastComplaint = 0; - if (readSource == XLOG_FROM_PG_WAL && emode == LOG) + /* use currentSource as readSource is reset at failure */ + if (currentSource == XLOG_FROM_PG_WAL && emode <= WARNING) { if (RecPtr == lastComplaint) emode = DEBUG1; diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c index f6446da2d6..78b85c5a25 100644 --- a/src/backend/replication/walreceiver.c +++ b/src/backend/replication/walreceiver.c @@ -498,10 +498,9 @@ WalReceiverMain(void) else if (len < 0) { ereport(LOG, - (errmsg("replication terminated by primary server"), - errdetail("End of WAL reached on timeline %u at %X/%X.", - startpointTLI, - LSN_FORMAT_ARGS(LogstreamResult.Write)))); + errmsg("replication terminated by primary server at %X/%X on timeline %u.", + LSN_FORMAT_ARGS(LogstreamResult.Write), + startpointTLI)); endofwal = true; break; } diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index 44b5c8726e..0f0cccd425 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -1276,9 +1276,16 @@ main(int argc, char **argv) exit(0); if (errormsg) - pg_fatal("error in WAL record at %X/%X: %s", - LSN_FORMAT_ARGS(xlogreader_state->ReadRecPtr), - errormsg); + { + if (xlogreader_state->EndOfWAL) + pg_log_info("end of WAL at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + else + pg_fatal("error in WAL record at %X/%X: %s", + LSN_FORMAT_ARGS(xlogreader_state->EndRecPtr), + errormsg); + } XLogReaderFree(xlogreader_state); diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h index d77bb2ab9b..84562d9de1 100644 --- a/src/include/access/xlogreader.h +++ b/src/include/access/xlogreader.h @@ -205,6 +205,7 @@ struct XLogReaderState */ XLogRecPtr ReadRecPtr; /* start of last record read */ XLogRecPtr EndRecPtr; /* end+1 of last record read */ + bool EndOfWAL; /* was the last attempt EOW? */ /* * Set at the end of recovery: the start point of a partial record at the diff --git a/src/test/recovery/t/035_recovery.pl b/src/test/recovery/t/035_recovery.pl new file mode 100644 index 0000000000..7107df6509 --- /dev/null +++ b/src/test/recovery/t/035_recovery.pl @@ -0,0 +1,130 @@ + +# Copyright (c) 2021-2023, PostgreSQL Global Development Group + +# Minimal test testing recovery process + +use strict; +use warnings; +use PostgreSQL::Test::Cluster; +use PostgreSQL::Test::Utils; +use Test::More; +use IPC::Run; + +my $reached_eow_pat = "reached end of WAL at "; +my $node = PostgreSQL::Test::Cluster->new('primary'); +$node->init(allows_streaming => 1); +$node->start; + +my ($stdout, $stderr) = ('', ''); + +my $segsize = $node->safe_psql('postgres', + qq[SELECT setting FROM pg_settings WHERE name = 'wal_segment_size';]); + +# make sure no records afterwards go to the next segment +$node->safe_psql('postgres', qq[ + SELECT pg_switch_wal(); + CHECKPOINT; + CREATE TABLE t(); +]); +$node->stop('immediate'); + +# identify REDO WAL file +my $cmd = "pg_controldata -D " . $node->data_dir(); +$cmd = ['pg_controldata', '-D', $node->data_dir()]; +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stdout =~ /^Latest checkpoint's REDO WAL file:[ \t] *(.+)$/m, + "checkpoint file is identified"); +my $chkptfile = $1; + +# identify the last record +my $walfile = $node->data_dir() . "/pg_wal/$chkptfile"; +$cmd = ['pg_waldump', $walfile]; +$stdout = ''; +$stderr = ''; +my $lastrec; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +foreach my $l (split(/\r?\n/, $stdout)) +{ + $lastrec = $l; +} +ok(defined $lastrec, "last WAL record is extracted"); +ok($stderr =~ /end of WAL at ([0-9A-F\/]+): .* at \g1/, + "pg_waldump emits the correct ending message"); + +# read the last record LSN excluding leading zeroes +ok ($lastrec =~ /, lsn: 0\/0*([1-9A-F][0-9A-F]+),/, + "LSN of the last record identified"); +my $lastlsn = $1; + +# corrupt the last record +my $offset = hex($lastlsn) % $segsize; +open(my $segf, '+<', $walfile) or die "failed to open $walfile\n"; +seek($segf, $offset, 0); # zero xl_tot_len, leaving following bytes alone. +print $segf "\0\0\0\0"; +close($segf); + +# pg_waldump complains about the corrupted record +$stdout = ''; +$stderr = ''; +IPC::Run::run $cmd, '>', \$stdout, '2>', \$stderr; +ok($stderr =~ /error: error in WAL record at 0\/$lastlsn: .* at 0\/$lastlsn/, + "pg_waldump emits the correct error message"); + +# also server complains for the same reason +my $logstart = -s $node->logfile; +$node->start; +ok($node->wait_for_log( + "WARNING: invalid record length at 0/$lastlsn: expected at least 24, got 0", + $logstart), + "header error is correctly logged at $lastlsn"); + +# no end-of-wal message should be seen this time +ok(!find_in_log($node, $reached_eow_pat, $logstart), + "false log message is not emitted"); + +# Create streaming standby linking to primary +my $backup_name = 'my_backup'; +$node->backup($backup_name); +my $node_standby = PostgreSQL::Test::Cluster->new('standby'); +$node_standby->init_from_backup($node, $backup_name, has_streaming => 1); +$node_standby->start; +$node->safe_psql('postgres', 'CREATE TABLE t ()'); +my $primary_lsn = $node->lsn('write'); +$node->wait_for_catchup($node_standby, 'write', $primary_lsn); + +$node_standby->stop(); +$node->stop('immediate'); + +# crash restart the primary +$logstart = -s $node->logfile; +$node->start(); +ok($node->wait_for_log($reached_eow_pat, $logstart), + 'primary properly emits end-of-WAL message'); + +# restart the standby +$logstart = -s $node_standby->logfile; +$node_standby->start(); +ok($node->wait_for_log($reached_eow_pat, $logstart), + 'standby properly emits end-of-WAL message'); + +$node_standby->stop(); +$node->stop(); + +done_testing(); + +#### helper routines +# find $pat in logfile of $node after $off-th byte +sub find_in_log +{ + my ($node, $pat, $off) = @_; + + $off = 0 unless defined $off; + my $log = PostgreSQL::Test::Utils::slurp_file($node->logfile); + return 0 if (length($log) <= $off); + + $log = substr($log, $off); + + return $log =~ m/$pat/; +} -- 2.31.1 ----Next_Part(Tue_Mar__7_15_35_47_2023_595)---- ^ permalink raw reply [nested|flat] 56+ messages in thread
* RE: Partial aggregates pushdown @ 2023-07-10 07:35 [email protected] <[email protected]> 2023-07-14 13:40 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]> 0 siblings, 1 reply; 56+ messages in thread From: [email protected] @ 2023-07-10 07:35 UTC (permalink / raw) To: Bruce Momjian <[email protected]>; Alexander Pyhalov <[email protected]>; pgsql-hackers; +Cc: Andres Freund <[email protected]>; Tom Lane <[email protected]>; Tomas Vondra <[email protected]>; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; Ilya Gladyshev <[email protected]>; [email protected] <[email protected]> Hi Mr.Bruce, Mr.Pyhalov, hackers. > From: Bruce Momjian <[email protected]> > Sent: Monday, June 12, 2023 10:38 PM > > On Mon, Jun 12, 2023 at 08:51:30AM +0000, [email protected] wrote: > > Hi Mr.Bruce, Mr.Pyhalov, hackers. > > > > Thank you for comments. I will try to respond to both of your comments as follows. > > I plan to start revising the patch next week. If you have any comments > > on the following respondences, I would appreciate it if you could give them to me this week. > > > > > From: Bruce Momjian <[email protected]> > > > Sent: Saturday, June 10, 2023 1:44 AM I agree that this feature is > > > designed for built-in sharding, but it is possible people could be > > > using aggregates on partitions backed by foreign tables without > > > sharding. Adding a requirement for non-sharding setups to need PG 17+ servers might be unreasonable. > > Indeed, it is possible to use partial aggregate pushdown feature for purposes other than sharding. > > The description of the section "F.38.6. Built-in sharding in > > PostgreSQL" assumes the use of Built-in sharding and will be modified to eliminate this assumption. > > The title of this section should be changed to something like "Aggregate on partitioned table". > > Sounds good. I have modified documents according to the above policy. > From: Bruce Momjian <[email protected]> > Sent: Thursday, June 22, 2023 8:39 PM > On Thu, Jun 22, 2023 at 05:23:33AM +0000, [email protected] wrote: > > Approach1-3: > > I will add a postgres_fdw option "check_partial_aggregate_support". > > This option is false, default. > > Only if this option is true, postgres_fdw connect to the remote server and get the version of the remote server. > > And if the version of the remote server is less than PG17, then partial aggregate push down to the remote server is > disable. > > Great! I have modified the program except for the point "if the version of the remote server is less than PG17". Instead, we have addressed the following. "If check_partial_aggregate_support is true and the remote server version is older than the local server version, postgres_fdw does not assume that the partial aggregate function is on the remote server unless the partial aggregate function and the aggregate function match." The reason for this is to maintain compatibility with any aggregate function that does not support partial aggregate in one version of V1 (V1 is PG17 or higher), even if the next version supports partial aggregate. For example, string_agg does not support partial aggregation in PG15, but it will support partial aggregation in PG16. We have not been able to add a test for the case where the remote server version is older than the local server version to the regression test. Is there any way to add such tests to the existing regression tests? Sincerely yours, Yuuki Fujii -- Yuuki Fujii Information Technology R&D Center Mitsubishi Electric Corporation Attachments: [application/octet-stream] 0001-Partial-aggregates-push-down-v23.patch (249.1K, ../../OS3PR01MB6660DF84634EDAFC55D7B0009530A@OS3PR01MB6660.jpnprd01.prod.outlook.com/2-0001-Partial-aggregates-push-down-v23.patch) download | inline diff: From: Yuki Fujii <[email protected]> Date: Mon, 10 Jul 2023 11:34:09 -0400 Subject: [PATCH] Partial aggregates push down v23 --- contrib/postgres_fdw/deparse.c | 100 ++- .../postgres_fdw/expected/postgres_fdw.out | 771 +++++++++++++++++- contrib/postgres_fdw/option.c | 4 +- contrib/postgres_fdw/postgres_fdw.c | 26 +- contrib/postgres_fdw/postgres_fdw.h | 9 + contrib/postgres_fdw/sql/postgres_fdw.sql | 294 ++++++- doc/src/sgml/catalogs.sgml | 13 + doc/src/sgml/postgres-fdw.sgml | 130 ++- doc/src/sgml/ref/create_aggregate.sgml | 42 + doc/src/sgml/xaggr.sgml | 15 +- src/backend/catalog/pg_aggregate.c | 101 +++ src/backend/commands/aggregatecmds.c | 4 + src/bin/pg_dump/pg_dump.c | 9 +- src/include/catalog/pg_aggregate.dat | 680 ++++++++++++--- src/include/catalog/pg_aggregate.h | 4 + src/include/catalog/pg_proc.dat | 196 +++++ .../regress/expected/create_aggregate.out | 184 +++++ src/test/regress/expected/oidjoins.out | 1 + src/test/regress/sql/create_aggregate.sql | 183 +++++ 19 files changed, 2587 insertions(+), 179 deletions(-) diff --git a/contrib/postgres_fdw/deparse.c b/contrib/postgres_fdw/deparse.c index 09d6dd60dd..a6cfc4a789 100644 --- a/contrib/postgres_fdw/deparse.c +++ b/contrib/postgres_fdw/deparse.c @@ -202,7 +202,7 @@ static bool is_subquery_var(Var *node, RelOptInfo *foreignrel, int *relno, int *colno); static void get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel, int *relno, int *colno); - +static bool partial_agg_ok(Aggref *agg, PgFdwRelationInfo *fpinfo); /* * Examine each qual clause in input_conds, and classify them into two groups, @@ -907,8 +907,9 @@ foreign_expr_walker(Node *node, if (!IS_UPPER_REL(glob_cxt->foreignrel)) return false; - /* Only non-split aggregates are pushable. */ - if (agg->aggsplit != AGGSPLIT_SIMPLE) + if ((agg->aggsplit != AGGSPLIT_SIMPLE) && (agg->aggsplit != AGGSPLIT_INITIAL_SERIAL)) + return false; + if (agg->aggsplit == AGGSPLIT_INITIAL_SERIAL && !partial_agg_ok(agg, fpinfo)) return false; /* As usual, it must be shippable. */ @@ -3517,14 +3518,34 @@ deparseAggref(Aggref *node, deparse_expr_cxt *context) StringInfo buf = context->buf; bool use_variadic; - /* Only basic, non-split aggregation accepted. */ - Assert(node->aggsplit == AGGSPLIT_SIMPLE); + + Assert((node->aggsplit == AGGSPLIT_SIMPLE) || + (node->aggsplit == AGGSPLIT_INITIAL_SERIAL)); /* Check if need to print VARIADIC (cf. ruleutils.c) */ use_variadic = node->aggvariadic; - /* Find aggregate name from aggfnoid which is a pg_proc entry */ - appendFunctionName(node->aggfnoid, context); + if (node->aggsplit == AGGSPLIT_SIMPLE) + /* Find aggregate name from aggfnoid which is a pg_proc entry */ + appendFunctionName(node->aggfnoid, context); + else + { + HeapTuple aggtup; + Form_pg_aggregate aggform; + + /* Find aggregate name from aggfnoid and aggpartialfn */ + aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(node->aggfnoid)); + if (!HeapTupleIsValid(aggtup)) + elog(ERROR, "cache lookup failed for aggregate %u", node->aggfnoid); + aggform = (Form_pg_aggregate) GETSTRUCT(aggtup); + + if (aggform->aggpartialfn) + appendFunctionName(aggform->aggpartialfn, context); + else + elog(ERROR, "there is no aggpartialfn %u", node->aggfnoid); + ReleaseSysCache(aggtup); + } + appendStringInfoChar(buf, '('); /* Add DISTINCT */ @@ -4047,3 +4068,68 @@ get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel, /* Shouldn't get here */ elog(ERROR, "unexpected expression in subquery output"); } + +/* + * Check that a buit-in aggpartialfunc exists on the remote server. If + * check_partial_aggregate_support is false, we assume the partial aggregate + * function exsits on the remote server. Otherwise we assume the partial + * aggregate function exsits on the remote server only if the remote server + * version is not less than the local server version. + */ +static bool +is_builtin_aggpartialfunc_shippable(Oid aggpartialfn, PgFdwRelationInfo *fpinfo) +{ + bool shippable = true; + + if (fpinfo->check_partial_aggregate_support) + { + if (fpinfo->remoteversion == 0) + { + PGconn *conn = GetConnection(fpinfo->user, false, NULL); + + fpinfo->remoteversion = PQserverVersion(conn); + } + if (fpinfo->remoteversion < PG_VERSION_NUM) + shippable = false; + } + return shippable; +} + +/* + * Check that partial aggregate agg is safe to push down. + * + * It is pushdown-safe when all of the following conditions are true. + * condition1) agg is AGGKIND_NORMAL aggregate which contains no distinct or + * order by clauses condition2) there is an aggregate function for partial + * aggregation (call it aggpartialfunc) corresponding to agg and + * aggpartialfunc exists on the remote server + */ +static bool +partial_agg_ok(Aggref *agg, PgFdwRelationInfo *fpinfo) +{ + HeapTuple aggtup; + Form_pg_aggregate aggform; + bool partial_agg_ok = true; + + Assert(agg->aggsplit == AGGSPLIT_INITIAL_SERIAL); + + /* We don't support complex partial aggregates */ + if (agg->aggdistinct || agg->aggkind != AGGKIND_NORMAL || agg->aggorder != NIL) + return false; + + aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(agg->aggfnoid)); + if (!HeapTupleIsValid(aggtup)) + elog(ERROR, "cache lookup failed for aggregate %u", agg->aggfnoid); + aggform = (Form_pg_aggregate) GETSTRUCT(aggtup); + + if (!aggform->aggpartialfn) + partial_agg_ok = false; + else if (is_builtin(aggform->aggpartialfn)) + partial_agg_ok = is_builtin_aggpartialfunc_shippable( + aggform->aggpartialfn, fpinfo); + else if (!is_shippable(aggform->aggpartialfn, ProcedureRelationId, fpinfo)) + partial_agg_ok = false; + + ReleaseSysCache(aggtup); + return partial_agg_ok; +} diff --git a/contrib/postgres_fdw/expected/postgres_fdw.out b/contrib/postgres_fdw/expected/postgres_fdw.out index c8c4614b54..8d80ba0a6b 100644 --- a/contrib/postgres_fdw/expected/postgres_fdw.out +++ b/contrib/postgres_fdw/expected/postgres_fdw.out @@ -9700,13 +9700,27 @@ RESET enable_partitionwise_join; -- =================================================================== -- test partitionwise aggregates -- =================================================================== -CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a); +ALTER SERVER loopback OPTIONS (ADD fdw_tuple_cost '0.5'); +CREATE TYPE mood AS enum ('sad', 'ok', 'happy'); +ALTER EXTENSION postgres_fdw ADD TYPE mood; +CREATE TABLE pagg_tab (a int, b int, c text, c_serial int4, + c_int4array _int4, c_interval interval, + c_money money, c_1c text, c_1b bytea, + c_bit bit(2), c_1or3int2 int2, + c_1or3int4 int4, c_1or3int8 int8, + c_bool bool, c_enum mood, c_pg_lsn pg_lsn, + c_tid tid, c_int4range int4range, + c_int4multirange int4multirange, + c_time time, c_timetz timetz, + c_timestamp timestamp, c_timestamptz timestamptz, + c_xid8 xid8) + PARTITION BY RANGE(a); CREATE TABLE pagg_tab_p1 (LIKE pagg_tab); CREATE TABLE pagg_tab_p2 (LIKE pagg_tab); CREATE TABLE pagg_tab_p3 (LIKE pagg_tab); -INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10; -INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10; -INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20; +INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i, array[(i % 2), 0], ((i % 2) || ' seconds')::interval, (i % 2)::money, ((i - 1) % 10)::text, (((i - 1) % 10)::text)::bytea, case when (i % 2) = 0 then B'01' else B'11' end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then true else false end, (case when (i % 2) = 0 then 'sad' else 'happy' end)::mood, ('0/' || i)::pg_lsn, ('(0,' || i || ')')::tid, int4range(0, i), int4multirange(int4range(0, i), int4range(100, 100 + i)), '00:00:00'::time + (i || ' seconds')::interval, '00:00:00'::timetz + (i || ' seconds')::interval, '2000-01-01'::timestamp + (i || ' seconds')::interval, '2000-01-01'::timestamptz + (i || ' seconds')::interval, ((i % 10)::text)::xid8 FROM generate_series(1, 3000) i WHERE (i % 30) < 10; +INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i, array[(i % 2), 0], ((i % 2) || ' seconds')::interval, (i % 2)::money, ((i - 1) % 10)::text, (((i - 1) % 10)::text)::bytea, case when (i % 2) = 0 then B'01' else B'11' end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then true else false end, (case when (i % 2) = 0 then 'sad' else 'happy' end)::mood, ('0/' || i)::pg_lsn, ('(0,' || i || ')')::tid, int4range(0, i), int4multirange(int4range(0, i), int4range(100, 100 + i)), '00:00:00'::time + (i || ' seconds')::interval, '00:00:00'::timetz + (i || ' seconds')::interval, '2000-01-01'::timestamp + (i || ' seconds')::interval, '2000-01-01'::timestamptz + (i || ' seconds')::interval, ((i % 10)::text)::xid8 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10; +INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i, array[(i % 2), 0], ((i % 2) || ' seconds')::interval, (i % 2)::money, ((i - 1) % 10)::text, (((i - 1) % 10)::text)::bytea, case when (i % 2) = 0 then B'01' else B'11' end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then true else false end, (case when (i % 2) = 0 then 'sad' else 'happy' end)::mood, ('0/' || i)::pg_lsn, ('(0,' || i || ')')::tid, int4range(0, i), int4multirange(int4range(0, i), int4range(100, 100 + i)), '00:00:00'::time + (i || ' seconds')::interval, '00:00:00'::timetz + (i || ' seconds')::interval, '2000-01-01'::timestamp + (i || ' seconds')::interval, '2000-01-01'::timestamptz + (i || ' seconds')::interval, ((i % 10)::text)::xid8 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20; -- Create foreign partitions CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1'); CREATE FOREIGN TABLE fpagg_tab_p2 PARTITION OF pagg_tab FOR VALUES FROM (10) TO (20) SERVER loopback OPTIONS (table_name 'pagg_tab_p2'); @@ -9765,8 +9779,8 @@ SELECT a, sum(b), min(b), count(*) FROM pagg_tab GROUP BY a HAVING avg(b) < 22 O -- Should have all the columns in the target list for the given relation EXPLAIN (VERBOSE, COSTS OFF) SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1; - QUERY PLAN ------------------------------------------------------------------------- + QUERY PLAN +------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Sort Output: t1.a, (count(((t1.*)::pagg_tab))) Sort Key: t1.a @@ -9777,21 +9791,21 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1; Filter: (avg(t1.b) < '22'::numeric) -> Foreign Scan on public.fpagg_tab_p1 t1 Output: t1.a, t1.*, t1.b - Remote SQL: SELECT a, b, c FROM public.pagg_tab_p1 + Remote SQL: SELECT a, b, c, c_serial, c_int4array, c_interval, c_money, c_1c, c_1b, c_bit, c_1or3int2, c_1or3int4, c_1or3int8, c_bool, c_enum, c_pg_lsn, c_tid, c_int4range, c_int4multirange, c_time, c_timetz, c_timestamp, c_timestamptz, c_xid8 FROM public.pagg_tab_p1 -> HashAggregate Output: t1_1.a, count(((t1_1.*)::pagg_tab)) Group Key: t1_1.a Filter: (avg(t1_1.b) < '22'::numeric) -> Foreign Scan on public.fpagg_tab_p2 t1_1 Output: t1_1.a, t1_1.*, t1_1.b - Remote SQL: SELECT a, b, c FROM public.pagg_tab_p2 + Remote SQL: SELECT a, b, c, c_serial, c_int4array, c_interval, c_money, c_1c, c_1b, c_bit, c_1or3int2, c_1or3int4, c_1or3int8, c_bool, c_enum, c_pg_lsn, c_tid, c_int4range, c_int4multirange, c_time, c_timetz, c_timestamp, c_timestamptz, c_xid8 FROM public.pagg_tab_p2 -> HashAggregate Output: t1_2.a, count(((t1_2.*)::pagg_tab)) Group Key: t1_2.a Filter: (avg(t1_2.b) < '22'::numeric) -> Foreign Scan on public.fpagg_tab_p3 t1_2 Output: t1_2.a, t1_2.*, t1_2.b - Remote SQL: SELECT a, b, c FROM public.pagg_tab_p3 + Remote SQL: SELECT a, b, c, c_serial, c_int4array, c_interval, c_money, c_1c, c_1b, c_bit, c_1or3int2, c_1or3int4, c_1or3int8, c_bool, c_enum, c_pg_lsn, c_tid, c_int4range, c_int4multirange, c_time, c_timetz, c_timestamp, c_timestamptz, c_xid8 FROM public.pagg_tab_p3 (25 rows) SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1; @@ -9810,11 +9824,11 @@ EXPLAIN (COSTS OFF) SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1; QUERY PLAN ----------------------------------------------------------------- - Sort - Sort Key: pagg_tab.b - -> Finalize HashAggregate - Group Key: pagg_tab.b - Filter: (sum(pagg_tab.a) < 700) + Finalize GroupAggregate + Group Key: pagg_tab.b + Filter: (sum(pagg_tab.a) < 700) + -> Sort + Sort Key: pagg_tab.b -> Append -> Partial HashAggregate Group Key: pagg_tab.b @@ -9827,6 +9841,735 @@ SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 -> Foreign Scan on fpagg_tab_p3 pagg_tab_2 (15 rows) +-- It's unsafe to push down having clause when there are partial aggregates +EXPLAIN (VERBOSE, COSTS OFF) +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1; + QUERY PLAN +--------------------------------------------------------------------------------------------------------------------------------------------- + Finalize GroupAggregate + Output: pagg_tab.b, avg(pagg_tab.a), max(pagg_tab.a), count(*) + Group Key: pagg_tab.b + Filter: (sum(pagg_tab.a) < 700) + -> Sort + Output: pagg_tab.b, (PARTIAL avg(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.a)) + Sort Key: pagg_tab.b + -> Append + -> Partial HashAggregate + Output: pagg_tab.b, PARTIAL avg(pagg_tab.a), PARTIAL max(pagg_tab.a), PARTIAL count(*), PARTIAL sum(pagg_tab.a) + Group Key: pagg_tab.b + -> Foreign Scan on public.fpagg_tab_p1 pagg_tab + Output: pagg_tab.b, pagg_tab.a + Remote SQL: SELECT a, b FROM public.pagg_tab_p1 + -> Partial HashAggregate + Output: pagg_tab_1.b, PARTIAL avg(pagg_tab_1.a), PARTIAL max(pagg_tab_1.a), PARTIAL count(*), PARTIAL sum(pagg_tab_1.a) + Group Key: pagg_tab_1.b + -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1 + Output: pagg_tab_1.b, pagg_tab_1.a + Remote SQL: SELECT a, b FROM public.pagg_tab_p2 + -> Partial HashAggregate + Output: pagg_tab_2.b, PARTIAL avg(pagg_tab_2.a), PARTIAL max(pagg_tab_2.a), PARTIAL count(*), PARTIAL sum(pagg_tab_2.a) + Group Key: pagg_tab_2.b + -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2 + Output: pagg_tab_2.b, pagg_tab_2.a + Remote SQL: SELECT a, b FROM public.pagg_tab_p3 +(26 rows) + +-- Partial aggregates are safe to push down without having clause +EXPLAIN (VERBOSE, COSTS OFF) +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; + QUERY PLAN +------------------------------------------------------------------------------------------------------------------------ + Sort + Output: pagg_tab.b, (avg(pagg_tab.a)), (max(pagg_tab.a)), (count(*)) + Sort Key: pagg_tab.b + -> Finalize HashAggregate + Output: pagg_tab.b, avg(pagg_tab.a), max(pagg_tab.a), count(*) + Group Key: pagg_tab.b + -> Append + -> Foreign Scan + Output: pagg_tab.b, (PARTIAL avg(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab) + Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p1 GROUP BY 1 + -> Foreign Scan + Output: pagg_tab_1.b, (PARTIAL avg(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1) + Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p2 GROUP BY 1 + -> Foreign Scan + Output: pagg_tab_2.b, (PARTIAL avg(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2) + Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p3 GROUP BY 1 +(19 rows) + +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; + b | avg | max | count +----+---------------------+-----+------- + 0 | 10.0000000000000000 | 20 | 60 + 1 | 11.0000000000000000 | 21 | 60 + 2 | 12.0000000000000000 | 22 | 60 + 3 | 13.0000000000000000 | 23 | 60 + 4 | 14.0000000000000000 | 24 | 60 + 5 | 15.0000000000000000 | 25 | 60 + 6 | 16.0000000000000000 | 26 | 60 + 7 | 17.0000000000000000 | 27 | 60 + 8 | 18.0000000000000000 | 28 | 60 + 9 | 19.0000000000000000 | 29 | 60 + 10 | 10.0000000000000000 | 20 | 60 + 11 | 11.0000000000000000 | 21 | 60 + 12 | 12.0000000000000000 | 22 | 60 + 13 | 13.0000000000000000 | 23 | 60 + 14 | 14.0000000000000000 | 24 | 60 + 15 | 15.0000000000000000 | 25 | 60 + 16 | 16.0000000000000000 | 26 | 60 + 17 | 17.0000000000000000 | 27 | 60 + 18 | 18.0000000000000000 | 28 | 60 + 19 | 19.0000000000000000 | 29 | 60 + 20 | 10.0000000000000000 | 20 | 60 + 21 | 11.0000000000000000 | 21 | 60 + 22 | 12.0000000000000000 | 22 | 60 + 23 | 13.0000000000000000 | 23 | 60 + 24 | 14.0000000000000000 | 24 | 60 + 25 | 15.0000000000000000 | 25 | 60 + 26 | 16.0000000000000000 | 26 | 60 + 27 | 17.0000000000000000 | 27 | 60 + 28 | 18.0000000000000000 | 28 | 60 + 29 | 19.0000000000000000 | 29 | 60 + 30 | 10.0000000000000000 | 20 | 60 + 31 | 11.0000000000000000 | 21 | 60 + 32 | 12.0000000000000000 | 22 | 60 + 33 | 13.0000000000000000 | 23 | 60 + 34 | 14.0000000000000000 | 24 | 60 + 35 | 15.0000000000000000 | 25 | 60 + 36 | 16.0000000000000000 | 26 | 60 + 37 | 17.0000000000000000 | 27 | 60 + 38 | 18.0000000000000000 | 28 | 60 + 39 | 19.0000000000000000 | 29 | 60 + 40 | 10.0000000000000000 | 20 | 60 + 41 | 11.0000000000000000 | 21 | 60 + 42 | 12.0000000000000000 | 22 | 60 + 43 | 13.0000000000000000 | 23 | 60 + 44 | 14.0000000000000000 | 24 | 60 + 45 | 15.0000000000000000 | 25 | 60 + 46 | 16.0000000000000000 | 26 | 60 + 47 | 17.0000000000000000 | 27 | 60 + 48 | 18.0000000000000000 | 28 | 60 + 49 | 19.0000000000000000 | 29 | 60 +(50 rows) + +-- Partial aggregates are safe to push down for all built-in aggregates +EXPLAIN (VERBOSE, COSTS OFF) +SELECT /* aggregate function <> aggpartialfunc */ + array_agg(c_int4array), array_agg(b), + avg(b::int2), avg(b::int4), avg(b::int8), avg(c_interval), avg(b::float4), avg(b::float8), avg(b::numeric), + corr(b::float8, (b * b)::float8), + covar_pop(b::float8, (b * b)::float8), + covar_samp(b::float8, (b * b)::float8), + regr_avgx((2 * b)::float8, b::float8), + regr_avgy((2 * b)::float8, b::float8), + regr_intercept((2 * b + 3)::float8, b::float8), + regr_r2((b * b)::float8, b::float8), + regr_slope((2 * b + 3)::float8, b::float8), + regr_sxx((2 * b + 3)::float8, b::float8), + regr_sxy((b * b)::float8, b::float8), + regr_syy((2 * b + 3)::float8, b::float8), + stddev(b::float4), stddev(b::float8), stddev(b::int2), stddev(b::int4), stddev(b::int8), stddev(b::numeric), + stddev_pop(b::float4), stddev_pop(b::float8), stddev_pop(b::int2), stddev_pop(b::int4), stddev_pop(b::int8), stddev_pop(b::numeric), + stddev_samp(b::float4), stddev_samp(b::float8), stddev_samp(b::int2), stddev_samp(b::int4), stddev_samp(b::int8), stddev_samp(b::numeric), + string_agg(c_1c, ','), string_agg(c_1b, ','), + sum(b::int8), sum(b::numeric), + variance(b::float4), variance(b::float8), variance(b::int2), variance(b::int4), variance(b::int8), variance(b::numeric), + var_pop(b::float4), var_pop(b::float8), var_pop(b::int2), var_pop(b::int4), var_pop(b::int8), var_pop(b::numeric), + var_samp(b::float4), var_samp(b::float8), var_samp(b::int2), var_samp(b::int4), var_samp(b::int8), var_samp(b::numeric), + /* aggregate function = aggpartialfunc */ + any_value(b * 0), + bit_and(c_bit), bit_and(c_1or3int2), bit_and(c_1or3int4), bit_and(c_1or3int8), + bit_or(c_bit), bit_or(c_1or3int2), bit_or(c_1or3int4), bit_or(c_1or3int8), + bit_xor(c_bit), bit_xor(c_1or3int2), bit_xor(c_1or3int4), bit_xor(c_1or3int8), + bool_and(c_bool), + bool_or(c_bool), + count(b), count(*), + every(c_bool), + max(c_int4array), max(c_enum), max(c_1c::char(1)), max('2000-01-01'::date + b), max('0.0.0.0'::inet + b), max(b::float4), max(b::float8), max(b::int2), max(b::int4), max(b::int8), max(c_interval), max(c_money), max(b::numeric), max(b::oid), max(c_pg_lsn), max(c_tid), max(c_1c), max(c_time), max(c_timetz), max(c_timestamp), max(c_timestamptz), max(c_xid8), + min(c_int4array), min(c_enum), min(c_1c::char(1)), min('2000-01-01'::date + b), min('0.0.0.0'::inet + b), min(b::float4), min(b::float8), min(b::int2), min(b::int4), min(b::int8), min(c_interval), min(c_money), min(b::numeric), min(b::oid), min(c_pg_lsn), min(c_tid), min(c_1c), min(c_time), min(c_timetz), min(c_timestamp), min(c_timestamptz), min(c_xid8), + range_intersect_agg(c_int4range), range_intersect_agg(c_int4multirange), + regr_count((2 * b + 3)::float8, b::float8), + sum(b::float4), sum(b::float8), sum(b::int2), sum(b::int4), sum(c_interval), sum(c_money) + FROM pagg_tab WHERE c_serial between 1 and 30; + QUERY PLAN +------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- + Finalize Aggregate + Output: array_agg(pagg_tab.c_int4array), array_agg(pagg_tab.b), avg((pagg_tab.b)::smallint), avg(pagg_tab.b), avg((pagg_tab.b)::bigint), avg(pagg_tab.c_interval), avg((pagg_tab.b)::real), avg((pagg_tab.b)::double precision), avg((pagg_tab.b)::numeric), corr((pagg_tab.b)::double precision, ((pagg_tab.b * pagg_tab.b))::double precision), covar_pop((pagg_tab.b)::double precision, ((pagg_tab.b * pagg_tab.b))::double precision), covar_samp((pagg_tab.b)::double precision, ((pagg_tab.b * pagg_tab.b))::double precision), regr_avgx(((2 * pagg_tab.b))::double precision, (pagg_tab.b)::double precision), regr_avgy(((2 * pagg_tab.b))::double precision, (pagg_tab.b)::double precision), regr_intercept((((2 * pagg_tab.b) + 3))::double precision, (pagg_tab.b)::double precision), regr_r2(((pagg_tab.b * pagg_tab.b))::double precision, (pagg_tab.b)::double precision), regr_slope((((2 * pagg_tab.b) + 3))::double precision, (pagg_tab.b)::double precision), regr_sxx((((2 * pagg_tab.b) + 3))::double precision, (pagg_tab.b)::double precision), regr_sxy(((pagg_tab.b * pagg_tab.b))::double precision, (pagg_tab.b)::double precision), regr_syy((((2 * pagg_tab.b) + 3))::double precision, (pagg_tab.b)::double precision), stddev((pagg_tab.b)::real), stddev((pagg_tab.b)::double precision), stddev((pagg_tab.b)::smallint), stddev(pagg_tab.b), stddev((pagg_tab.b)::bigint), stddev((pagg_tab.b)::numeric), stddev_pop((pagg_tab.b)::real), stddev_pop((pagg_tab.b)::double precision), stddev_pop((pagg_tab.b)::smallint), stddev_pop(pagg_tab.b), stddev_pop((pagg_tab.b)::bigint), stddev_pop((pagg_tab.b)::numeric), stddev_samp((pagg_tab.b)::real), stddev_samp((pagg_tab.b)::double precision), stddev_samp((pagg_tab.b)::smallint), stddev_samp(pagg_tab.b), stddev_samp((pagg_tab.b)::bigint), stddev_samp((pagg_tab.b)::numeric), string_agg(pagg_tab.c_1c, ','::text), string_agg(pagg_tab.c_1b, '\x2c'::bytea), sum((pagg_tab.b)::bigint), sum((pagg_tab.b)::numeric), variance((pagg_tab.b)::real), variance((pagg_tab.b)::double precision), variance((pagg_tab.b)::smallint), variance(pagg_tab.b), variance((pagg_tab.b)::bigint), variance((pagg_tab.b)::numeric), var_pop((pagg_tab.b)::real), var_pop((pagg_tab.b)::double precision), var_pop((pagg_tab.b)::smallint), var_pop(pagg_tab.b), var_pop((pagg_tab.b)::bigint), var_pop((pagg_tab.b)::numeric), var_samp((pagg_tab.b)::real), var_samp((pagg_tab.b)::double precision), var_samp((pagg_tab.b)::smallint), var_samp(pagg_tab.b), var_samp((pagg_tab.b)::bigint), var_samp((pagg_tab.b)::numeric), any_value((pagg_tab.b * 0)), bit_and(pagg_tab.c_bit), bit_and(pagg_tab.c_1or3int2), bit_and(pagg_tab.c_1or3int4), bit_and(pagg_tab.c_1or3int8), bit_or(pagg_tab.c_bit), bit_or(pagg_tab.c_1or3int2), bit_or(pagg_tab.c_1or3int4), bit_or(pagg_tab.c_1or3int8), bit_xor(pagg_tab.c_bit), bit_xor(pagg_tab.c_1or3int2), bit_xor(pagg_tab.c_1or3int4), bit_xor(pagg_tab.c_1or3int8), bool_and(pagg_tab.c_bool), bool_or(pagg_tab.c_bool), count(pagg_tab.b), count(*), every(pagg_tab.c_bool), max(pagg_tab.c_int4array), max(pagg_tab.c_enum), max((pagg_tab.c_1c)::character(1)), max(('01-01-2000'::date + pagg_tab.b)), max(('0.0.0.0'::inet + (pagg_tab.b)::bigint)), max((pagg_tab.b)::real), max((pagg_tab.b)::double precision), max((pagg_tab.b)::smallint), max(pagg_tab.b), max((pagg_tab.b)::bigint), max(pagg_tab.c_interval), max(pagg_tab.c_money), max((pagg_tab.b)::numeric), max((pagg_tab.b)::oid), max(pagg_tab.c_pg_lsn), max(pagg_tab.c_tid), max(pagg_tab.c_1c), max(pagg_tab.c_time), max(pagg_tab.c_timetz), max(pagg_tab.c_timestamp), max(pagg_tab.c_timestamptz), max(pagg_tab.c_xid8), min(pagg_tab.c_int4array), min(pagg_tab.c_enum), min((pagg_tab.c_1c)::character(1)), min(('01-01-2000'::date + pagg_tab.b)), min(('0.0.0.0'::inet + (pagg_tab.b)::bigint)), min((pagg_tab.b)::real), min((pagg_tab.b)::double precision), min((pagg_tab.b)::smallint), min(pagg_tab.b), min((pagg_tab.b)::bigint), min(pagg_tab.c_interval), min(pagg_tab.c_money), min((pagg_tab.b)::numeric), min((pagg_tab.b)::oid), min(pagg_tab.c_pg_lsn), min(pagg_tab.c_tid), min(pagg_tab.c_1c), min(pagg_tab.c_time), min(pagg_tab.c_timetz), min(pagg_tab.c_timestamp), min(pagg_tab.c_timestamptz), min(pagg_tab.c_xid8), range_intersect_agg(pagg_tab.c_int4range), range_intersect_agg(pagg_tab.c_int4multirange), regr_count((((2 * pagg_tab.b) + 3))::double precision, (pagg_tab.b)::double precision), sum((pagg_tab.b)::real), sum((pagg_tab.b)::double precision), sum((pagg_tab.b)::smallint), sum(pagg_tab.b), sum(pagg_tab.c_interval), sum(pagg_tab.c_money) + -> Append + -> Foreign Scan + Output: (PARTIAL array_agg(pagg_tab.c_int4array)), (PARTIAL array_agg(pagg_tab.b)), (PARTIAL avg((pagg_tab.b)::smallint)), (PARTIAL avg(pagg_tab.b)), (PARTIAL avg((pagg_tab.b)::bigint)), (PARTIAL avg(pagg_tab.c_interval)), (PARTIAL avg((pagg_tab.b)::real)), (PARTIAL avg((pagg_tab.b)::double precision)), (PARTIAL avg((pagg_tab.b)::numeric)), (PARTIAL corr((pagg_tab.b)::double precision, ((pagg_tab.b * pagg_tab.b))::double precision)), (PARTIAL covar_pop((pagg_tab.b)::double precision, ((pagg_tab.b * pagg_tab.b))::double precision)), (PARTIAL covar_samp((pagg_tab.b)::double precision, ((pagg_tab.b * pagg_tab.b))::double precision)), (PARTIAL regr_avgx(((2 * pagg_tab.b))::double precision, (pagg_tab.b)::double precision)), (PARTIAL regr_avgy(((2 * pagg_tab.b))::double precision, (pagg_tab.b)::double precision)), (PARTIAL regr_intercept((((2 * pagg_tab.b) + 3))::double precision, (pagg_tab.b)::double precision)), (PARTIAL regr_r2(((pagg_tab.b * pagg_tab.b))::double precision, (pagg_tab.b)::double precision)), (PARTIAL regr_slope((((2 * pagg_tab.b) + 3))::double precision, (pagg_tab.b)::double precision)), (PARTIAL regr_sxx((((2 * pagg_tab.b) + 3))::double precision, (pagg_tab.b)::double precision)), (PARTIAL regr_sxy(((pagg_tab.b * pagg_tab.b))::double precision, (pagg_tab.b)::double precision)), (PARTIAL regr_syy((((2 * pagg_tab.b) + 3))::double precision, (pagg_tab.b)::double precision)), (PARTIAL stddev((pagg_tab.b)::real)), (PARTIAL stddev((pagg_tab.b)::double precision)), (PARTIAL stddev((pagg_tab.b)::smallint)), (PARTIAL stddev(pagg_tab.b)), (PARTIAL stddev((pagg_tab.b)::bigint)), (PARTIAL stddev((pagg_tab.b)::numeric)), (PARTIAL stddev_pop((pagg_tab.b)::real)), (PARTIAL stddev_pop((pagg_tab.b)::double precision)), (PARTIAL stddev_pop((pagg_tab.b)::smallint)), (PARTIAL stddev_pop(pagg_tab.b)), (PARTIAL stddev_pop((pagg_tab.b)::bigint)), (PARTIAL stddev_pop((pagg_tab.b)::numeric)), (PARTIAL stddev_samp((pagg_tab.b)::real)), (PARTIAL stddev_samp((pagg_tab.b)::double precision)), (PARTIAL stddev_samp((pagg_tab.b)::smallint)), (PARTIAL stddev_samp(pagg_tab.b)), (PARTIAL stddev_samp((pagg_tab.b)::bigint)), (PARTIAL stddev_samp((pagg_tab.b)::numeric)), (PARTIAL string_agg(pagg_tab.c_1c, ','::text)), (PARTIAL string_agg(pagg_tab.c_1b, '\x2c'::bytea)), (PARTIAL sum((pagg_tab.b)::bigint)), (PARTIAL sum((pagg_tab.b)::numeric)), (PARTIAL variance((pagg_tab.b)::real)), (PARTIAL variance((pagg_tab.b)::double precision)), (PARTIAL variance((pagg_tab.b)::smallint)), (PARTIAL variance(pagg_tab.b)), (PARTIAL variance((pagg_tab.b)::bigint)), (PARTIAL variance((pagg_tab.b)::numeric)), (PARTIAL var_pop((pagg_tab.b)::real)), (PARTIAL var_pop((pagg_tab.b)::double precision)), (PARTIAL var_pop((pagg_tab.b)::smallint)), (PARTIAL var_pop(pagg_tab.b)), (PARTIAL var_pop((pagg_tab.b)::bigint)), (PARTIAL var_pop((pagg_tab.b)::numeric)), (PARTIAL var_samp((pagg_tab.b)::real)), (PARTIAL var_samp((pagg_tab.b)::double precision)), (PARTIAL var_samp((pagg_tab.b)::smallint)), (PARTIAL var_samp(pagg_tab.b)), (PARTIAL var_samp((pagg_tab.b)::bigint)), (PARTIAL var_samp((pagg_tab.b)::numeric)), (PARTIAL any_value((pagg_tab.b * 0))), (PARTIAL bit_and(pagg_tab.c_bit)), (PARTIAL bit_and(pagg_tab.c_1or3int2)), (PARTIAL bit_and(pagg_tab.c_1or3int4)), (PARTIAL bit_and(pagg_tab.c_1or3int8)), (PARTIAL bit_or(pagg_tab.c_bit)), (PARTIAL bit_or(pagg_tab.c_1or3int2)), (PARTIAL bit_or(pagg_tab.c_1or3int4)), (PARTIAL bit_or(pagg_tab.c_1or3int8)), (PARTIAL bit_xor(pagg_tab.c_bit)), (PARTIAL bit_xor(pagg_tab.c_1or3int2)), (PARTIAL bit_xor(pagg_tab.c_1or3int4)), (PARTIAL bit_xor(pagg_tab.c_1or3int8)), (PARTIAL bool_and(pagg_tab.c_bool)), (PARTIAL bool_or(pagg_tab.c_bool)), (PARTIAL count(pagg_tab.b)), (PARTIAL count(*)), (PARTIAL every(pagg_tab.c_bool)), (PARTIAL max(pagg_tab.c_int4array)), (PARTIAL max(pagg_tab.c_enum)), (PARTIAL max((pagg_tab.c_1c)::character(1))), (PARTIAL max(('01-01-2000'::date + pagg_tab.b))), (PARTIAL max(('0.0.0.0'::inet + (pagg_tab.b)::bigint))), (PARTIAL max((pagg_tab.b)::real)), (PARTIAL max((pagg_tab.b)::double precision)), (PARTIAL max((pagg_tab.b)::smallint)), (PARTIAL max(pagg_tab.b)), (PARTIAL max((pagg_tab.b)::bigint)), (PARTIAL max(pagg_tab.c_interval)), (PARTIAL max(pagg_tab.c_money)), (PARTIAL max((pagg_tab.b)::numeric)), (PARTIAL max((pagg_tab.b)::oid)), (PARTIAL max(pagg_tab.c_pg_lsn)), (PARTIAL max(pagg_tab.c_tid)), (PARTIAL max(pagg_tab.c_1c)), (PARTIAL max(pagg_tab.c_time)), (PARTIAL max(pagg_tab.c_timetz)), (PARTIAL max(pagg_tab.c_timestamp)), (PARTIAL max(pagg_tab.c_timestamptz)), (PARTIAL max(pagg_tab.c_xid8)), (PARTIAL min(pagg_tab.c_int4array)), (PARTIAL min(pagg_tab.c_enum)), (PARTIAL min((pagg_tab.c_1c)::character(1))), (PARTIAL min(('01-01-2000'::date + pagg_tab.b))), (PARTIAL min(('0.0.0.0'::inet + (pagg_tab.b)::bigint))), (PARTIAL min((pagg_tab.b)::real)), (PARTIAL min((pagg_tab.b)::double precision)), (PARTIAL min((pagg_tab.b)::smallint)), (PARTIAL min(pagg_tab.b)), (PARTIAL min((pagg_tab.b)::bigint)), (PARTIAL min(pagg_tab.c_interval)), (PARTIAL min(pagg_tab.c_money)), (PARTIAL min((pagg_tab.b)::numeric)), (PARTIAL min((pagg_tab.b)::oid)), (PARTIAL min(pagg_tab.c_pg_lsn)), (PARTIAL min(pagg_tab.c_tid)), (PARTIAL min(pagg_tab.c_1c)), (PARTIAL min(pagg_tab.c_time)), (PARTIAL min(pagg_tab.c_timetz)), (PARTIAL min(pagg_tab.c_timestamp)), (PARTIAL min(pagg_tab.c_timestamptz)), (PARTIAL min(pagg_tab.c_xid8)), (PARTIAL range_intersect_agg(pagg_tab.c_int4range)), (PARTIAL range_intersect_agg(pagg_tab.c_int4multirange)), (PARTIAL regr_count((((2 * pagg_tab.b) + 3))::double precision, (pagg_tab.b)::double precision)), (PARTIAL sum((pagg_tab.b)::real)), (PARTIAL sum((pagg_tab.b)::double precision)), (PARTIAL sum((pagg_tab.b)::smallint)), (PARTIAL sum(pagg_tab.b)), (PARTIAL sum(pagg_tab.c_interval)), (PARTIAL sum(pagg_tab.c_money)) + Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab) + Remote SQL: SELECT array_agg_p_anyarray(c_int4array), array_agg_p_anynonarray(b), avg_p_int2(b::smallint), avg_p_int4(b), avg_p_int8(b::bigint), avg_p_interval(c_interval), avg_p_float4(b::real), avg_p_float8(b::double precision), avg_p_numeric(b::numeric), corr_p(b::double precision, (b * b)::double precision), covar_pop_p(b::double precision, (b * b)::double precision), covar_samp_p(b::double precision, (b * b)::double precision), regr_avgx_p((2 * b)::double precision, b::double precision), regr_avgy_p((2 * b)::double precision, b::double precision), regr_intercept_p(((2 * b) + 3)::double precision, b::double precision), regr_r2_p((b * b)::double precision, b::double precision), regr_slope_p(((2 * b) + 3)::double precision, b::double precision), regr_sxx_p(((2 * b) + 3)::double precision, b::double precision), regr_sxy_p((b * b)::double precision, b::double precision), regr_syy_p(((2 * b) + 3)::double precision, b::double precision), stddev_p_float4(b::real), stddev_p_float8(b::double precision), stddev_p_int2(b::smallint), stddev_p_int4(b), stddev_p_int8(b::bigint), stddev_p_numeric(b::numeric), stddev_pop_p_float4(b::real), stddev_pop_p_float8(b::double precision), stddev_pop_p_int2(b::smallint), stddev_pop_p_int4(b), stddev_pop_p_int8(b::bigint), stddev_pop_p_numeric(b::numeric), stddev_samp_p_float4(b::real), stddev_samp_p_float8(b::double precision), stddev_samp_p_int2(b::smallint), stddev_samp_p_int4(b), stddev_samp_p_int8(b::bigint), stddev_samp_p_numeric(b::numeric), string_agg_p_text_text(c_1c, ','::text), string_agg_p_bytea_bytea(c_1b, E'\\x2c'::bytea), sum_p_int8(b::bigint), sum_p_numeric(b::numeric), variance_p_float4(b::real), variance_p_float8(b::double precision), variance_p_int2(b::smallint), variance_p_int4(b), variance_p_int8(b::bigint), variance_p_numeric(b::numeric), var_pop_p_float4(b::real), var_pop_p_float8(b::double precision), var_pop_p_int2(b::smallint), var_pop_p_int4(b), var_pop_p_int8(b::bigint), var_pop_p_numeric(b::numeric), var_samp_p_float4(b::real), var_samp_p_float8(b::double precision), var_samp_p_int2(b::smallint), var_samp_p_int4(b), var_samp_p_int8(b::bigint), var_samp_p_numeric(b::numeric), any_value((b * 0)), bit_and(c_bit), bit_and(c_1or3int2), bit_and(c_1or3int4), bit_and(c_1or3int8), bit_or(c_bit), bit_or(c_1or3int2), bit_or(c_1or3int4), bit_or(c_1or3int8), bit_xor(c_bit), bit_xor(c_1or3int2), bit_xor(c_1or3int4), bit_xor(c_1or3int8), bool_and(c_bool), bool_or(c_bool), count(b), count(*), every(c_bool), max(c_int4array), max(c_enum), max(c_1c::character(1)), max(('01-01-2000'::date + b)), max(('0.0.0.0'::inet + b)), max(b::real), max(b::double precision), max(b::smallint), max(b), max(b::bigint), max(c_interval), max(c_money), max(b::numeric), max(b::oid), max(c_pg_lsn), max(c_tid), max(c_1c), max(c_time), max(c_timetz), max(c_timestamp), max(c_timestamptz), max(c_xid8), min(c_int4array), min(c_enum), min(c_1c::character(1)), min(('01-01-2000'::date + b)), min(('0.0.0.0'::inet + b)), min(b::real), min(b::double precision), min(b::smallint), min(b), min(b::bigint), min(c_interval), min(c_money), min(b::numeric), min(b::oid), min(c_pg_lsn), min(c_tid), min(c_1c), min(c_time), min(c_timetz), min(c_timestamp), min(c_timestamptz), min(c_xid8), range_intersect_agg(c_int4range), range_intersect_agg(c_int4multirange), regr_count(((2 * b) + 3)::double precision, b::double precision), sum(b::real), sum(b::double precision), sum(b::smallint), sum(b), sum(c_interval), sum(c_money) FROM public.pagg_tab_p1 WHERE ((c_serial >= 1)) AND ((c_serial <= 30)) + -> Foreign Scan + Output: (PARTIAL array_agg(pagg_tab_1.c_int4array)), (PARTIAL array_agg(pagg_tab_1.b)), (PARTIAL avg((pagg_tab_1.b)::smallint)), (PARTIAL avg(pagg_tab_1.b)), (PARTIAL avg((pagg_tab_1.b)::bigint)), (PARTIAL avg(pagg_tab_1.c_interval)), (PARTIAL avg((pagg_tab_1.b)::real)), (PARTIAL avg((pagg_tab_1.b)::double precision)), (PARTIAL avg((pagg_tab_1.b)::numeric)), (PARTIAL corr((pagg_tab_1.b)::double precision, ((pagg_tab_1.b * pagg_tab_1.b))::double precision)), (PARTIAL covar_pop((pagg_tab_1.b)::double precision, ((pagg_tab_1.b * pagg_tab_1.b))::double precision)), (PARTIAL covar_samp((pagg_tab_1.b)::double precision, ((pagg_tab_1.b * pagg_tab_1.b))::double precision)), (PARTIAL regr_avgx(((2 * pagg_tab_1.b))::double precision, (pagg_tab_1.b)::double precision)), (PARTIAL regr_avgy(((2 * pagg_tab_1.b))::double precision, (pagg_tab_1.b)::double precision)), (PARTIAL regr_intercept((((2 * pagg_tab_1.b) + 3))::double precision, (pagg_tab_1.b)::double precision)), (PARTIAL regr_r2(((pagg_tab_1.b * pagg_tab_1.b))::double precision, (pagg_tab_1.b)::double precision)), (PARTIAL regr_slope((((2 * pagg_tab_1.b) + 3))::double precision, (pagg_tab_1.b)::double precision)), (PARTIAL regr_sxx((((2 * pagg_tab_1.b) + 3))::double precision, (pagg_tab_1.b)::double precision)), (PARTIAL regr_sxy(((pagg_tab_1.b * pagg_tab_1.b))::double precision, (pagg_tab_1.b)::double precision)), (PARTIAL regr_syy((((2 * pagg_tab_1.b) + 3))::double precision, (pagg_tab_1.b)::double precision)), (PARTIAL stddev((pagg_tab_1.b)::real)), (PARTIAL stddev((pagg_tab_1.b)::double precision)), (PARTIAL stddev((pagg_tab_1.b)::smallint)), (PARTIAL stddev(pagg_tab_1.b)), (PARTIAL stddev((pagg_tab_1.b)::bigint)), (PARTIAL stddev((pagg_tab_1.b)::numeric)), (PARTIAL stddev_pop((pagg_tab_1.b)::real)), (PARTIAL stddev_pop((pagg_tab_1.b)::double precision)), (PARTIAL stddev_pop((pagg_tab_1.b)::smallint)), (PARTIAL stddev_pop(pagg_tab_1.b)), (PARTIAL stddev_pop((pagg_tab_1.b)::bigint)), (PARTIAL stddev_pop((pagg_tab_1.b)::numeric)), (PARTIAL stddev_samp((pagg_tab_1.b)::real)), (PARTIAL stddev_samp((pagg_tab_1.b)::double precision)), (PARTIAL stddev_samp((pagg_tab_1.b)::smallint)), (PARTIAL stddev_samp(pagg_tab_1.b)), (PARTIAL stddev_samp((pagg_tab_1.b)::bigint)), (PARTIAL stddev_samp((pagg_tab_1.b)::numeric)), (PARTIAL string_agg(pagg_tab_1.c_1c, ','::text)), (PARTIAL string_agg(pagg_tab_1.c_1b, '\x2c'::bytea)), (PARTIAL sum((pagg_tab_1.b)::bigint)), (PARTIAL sum((pagg_tab_1.b)::numeric)), (PARTIAL variance((pagg_tab_1.b)::real)), (PARTIAL variance((pagg_tab_1.b)::double precision)), (PARTIAL variance((pagg_tab_1.b)::smallint)), (PARTIAL variance(pagg_tab_1.b)), (PARTIAL variance((pagg_tab_1.b)::bigint)), (PARTIAL variance((pagg_tab_1.b)::numeric)), (PARTIAL var_pop((pagg_tab_1.b)::real)), (PARTIAL var_pop((pagg_tab_1.b)::double precision)), (PARTIAL var_pop((pagg_tab_1.b)::smallint)), (PARTIAL var_pop(pagg_tab_1.b)), (PARTIAL var_pop((pagg_tab_1.b)::bigint)), (PARTIAL var_pop((pagg_tab_1.b)::numeric)), (PARTIAL var_samp((pagg_tab_1.b)::real)), (PARTIAL var_samp((pagg_tab_1.b)::double precision)), (PARTIAL var_samp((pagg_tab_1.b)::smallint)), (PARTIAL var_samp(pagg_tab_1.b)), (PARTIAL var_samp((pagg_tab_1.b)::bigint)), (PARTIAL var_samp((pagg_tab_1.b)::numeric)), (PARTIAL any_value((pagg_tab_1.b * 0))), (PARTIAL bit_and(pagg_tab_1.c_bit)), (PARTIAL bit_and(pagg_tab_1.c_1or3int2)), (PARTIAL bit_and(pagg_tab_1.c_1or3int4)), (PARTIAL bit_and(pagg_tab_1.c_1or3int8)), (PARTIAL bit_or(pagg_tab_1.c_bit)), (PARTIAL bit_or(pagg_tab_1.c_1or3int2)), (PARTIAL bit_or(pagg_tab_1.c_1or3int4)), (PARTIAL bit_or(pagg_tab_1.c_1or3int8)), (PARTIAL bit_xor(pagg_tab_1.c_bit)), (PARTIAL bit_xor(pagg_tab_1.c_1or3int2)), (PARTIAL bit_xor(pagg_tab_1.c_1or3int4)), (PARTIAL bit_xor(pagg_tab_1.c_1or3int8)), (PARTIAL bool_and(pagg_tab_1.c_bool)), (PARTIAL bool_or(pagg_tab_1.c_bool)), (PARTIAL count(pagg_tab_1.b)), (PARTIAL count(*)), (PARTIAL every(pagg_tab_1.c_bool)), (PARTIAL max(pagg_tab_1.c_int4array)), (PARTIAL max(pagg_tab_1.c_enum)), (PARTIAL max((pagg_tab_1.c_1c)::character(1))), (PARTIAL max(('01-01-2000'::date + pagg_tab_1.b))), (PARTIAL max(('0.0.0.0'::inet + (pagg_tab_1.b)::bigint))), (PARTIAL max((pagg_tab_1.b)::real)), (PARTIAL max((pagg_tab_1.b)::double precision)), (PARTIAL max((pagg_tab_1.b)::smallint)), (PARTIAL max(pagg_tab_1.b)), (PARTIAL max((pagg_tab_1.b)::bigint)), (PARTIAL max(pagg_tab_1.c_interval)), (PARTIAL max(pagg_tab_1.c_money)), (PARTIAL max((pagg_tab_1.b)::numeric)), (PARTIAL max((pagg_tab_1.b)::oid)), (PARTIAL max(pagg_tab_1.c_pg_lsn)), (PARTIAL max(pagg_tab_1.c_tid)), (PARTIAL max(pagg_tab_1.c_1c)), (PARTIAL max(pagg_tab_1.c_time)), (PARTIAL max(pagg_tab_1.c_timetz)), (PARTIAL max(pagg_tab_1.c_timestamp)), (PARTIAL max(pagg_tab_1.c_timestamptz)), (PARTIAL max(pagg_tab_1.c_xid8)), (PARTIAL min(pagg_tab_1.c_int4array)), (PARTIAL min(pagg_tab_1.c_enum)), (PARTIAL min((pagg_tab_1.c_1c)::character(1))), (PARTIAL min(('01-01-2000'::date + pagg_tab_1.b))), (PARTIAL min(('0.0.0.0'::inet + (pagg_tab_1.b)::bigint))), (PARTIAL min((pagg_tab_1.b)::real)), (PARTIAL min((pagg_tab_1.b)::double precision)), (PARTIAL min((pagg_tab_1.b)::smallint)), (PARTIAL min(pagg_tab_1.b)), (PARTIAL min((pagg_tab_1.b)::bigint)), (PARTIAL min(pagg_tab_1.c_interval)), (PARTIAL min(pagg_tab_1.c_money)), (PARTIAL min((pagg_tab_1.b)::numeric)), (PARTIAL min((pagg_tab_1.b)::oid)), (PARTIAL min(pagg_tab_1.c_pg_lsn)), (PARTIAL min(pagg_tab_1.c_tid)), (PARTIAL min(pagg_tab_1.c_1c)), (PARTIAL min(pagg_tab_1.c_time)), (PARTIAL min(pagg_tab_1.c_timetz)), (PARTIAL min(pagg_tab_1.c_timestamp)), (PARTIAL min(pagg_tab_1.c_timestamptz)), (PARTIAL min(pagg_tab_1.c_xid8)), (PARTIAL range_intersect_agg(pagg_tab_1.c_int4range)), (PARTIAL range_intersect_agg(pagg_tab_1.c_int4multirange)), (PARTIAL regr_count((((2 * pagg_tab_1.b) + 3))::double precision, (pagg_tab_1.b)::double precision)), (PARTIAL sum((pagg_tab_1.b)::real)), (PARTIAL sum((pagg_tab_1.b)::double precision)), (PARTIAL sum((pagg_tab_1.b)::smallint)), (PARTIAL sum(pagg_tab_1.b)), (PARTIAL sum(pagg_tab_1.c_interval)), (PARTIAL sum(pagg_tab_1.c_money)) + Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1) + Remote SQL: SELECT array_agg_p_anyarray(c_int4array), array_agg_p_anynonarray(b), avg_p_int2(b::smallint), avg_p_int4(b), avg_p_int8(b::bigint), avg_p_interval(c_interval), avg_p_float4(b::real), avg_p_float8(b::double precision), avg_p_numeric(b::numeric), corr_p(b::double precision, (b * b)::double precision), covar_pop_p(b::double precision, (b * b)::double precision), covar_samp_p(b::double precision, (b * b)::double precision), regr_avgx_p((2 * b)::double precision, b::double precision), regr_avgy_p((2 * b)::double precision, b::double precision), regr_intercept_p(((2 * b) + 3)::double precision, b::double precision), regr_r2_p((b * b)::double precision, b::double precision), regr_slope_p(((2 * b) + 3)::double precision, b::double precision), regr_sxx_p(((2 * b) + 3)::double precision, b::double precision), regr_sxy_p((b * b)::double precision, b::double precision), regr_syy_p(((2 * b) + 3)::double precision, b::double precision), stddev_p_float4(b::real), stddev_p_float8(b::double precision), stddev_p_int2(b::smallint), stddev_p_int4(b), stddev_p_int8(b::bigint), stddev_p_numeric(b::numeric), stddev_pop_p_float4(b::real), stddev_pop_p_float8(b::double precision), stddev_pop_p_int2(b::smallint), stddev_pop_p_int4(b), stddev_pop_p_int8(b::bigint), stddev_pop_p_numeric(b::numeric), stddev_samp_p_float4(b::real), stddev_samp_p_float8(b::double precision), stddev_samp_p_int2(b::smallint), stddev_samp_p_int4(b), stddev_samp_p_int8(b::bigint), stddev_samp_p_numeric(b::numeric), string_agg_p_text_text(c_1c, ','::text), string_agg_p_bytea_bytea(c_1b, E'\\x2c'::bytea), sum_p_int8(b::bigint), sum_p_numeric(b::numeric), variance_p_float4(b::real), variance_p_float8(b::double precision), variance_p_int2(b::smallint), variance_p_int4(b), variance_p_int8(b::bigint), variance_p_numeric(b::numeric), var_pop_p_float4(b::real), var_pop_p_float8(b::double precision), var_pop_p_int2(b::smallint), var_pop_p_int4(b), var_pop_p_int8(b::bigint), var_pop_p_numeric(b::numeric), var_samp_p_float4(b::real), var_samp_p_float8(b::double precision), var_samp_p_int2(b::smallint), var_samp_p_int4(b), var_samp_p_int8(b::bigint), var_samp_p_numeric(b::numeric), any_value((b * 0)), bit_and(c_bit), bit_and(c_1or3int2), bit_and(c_1or3int4), bit_and(c_1or3int8), bit_or(c_bit), bit_or(c_1or3int2), bit_or(c_1or3int4), bit_or(c_1or3int8), bit_xor(c_bit), bit_xor(c_1or3int2), bit_xor(c_1or3int4), bit_xor(c_1or3int8), bool_and(c_bool), bool_or(c_bool), count(b), count(*), every(c_bool), max(c_int4array), max(c_enum), max(c_1c::character(1)), max(('01-01-2000'::date + b)), max(('0.0.0.0'::inet + b)), max(b::real), max(b::double precision), max(b::smallint), max(b), max(b::bigint), max(c_interval), max(c_money), max(b::numeric), max(b::oid), max(c_pg_lsn), max(c_tid), max(c_1c), max(c_time), max(c_timetz), max(c_timestamp), max(c_timestamptz), max(c_xid8), min(c_int4array), min(c_enum), min(c_1c::character(1)), min(('01-01-2000'::date + b)), min(('0.0.0.0'::inet + b)), min(b::real), min(b::double precision), min(b::smallint), min(b), min(b::bigint), min(c_interval), min(c_money), min(b::numeric), min(b::oid), min(c_pg_lsn), min(c_tid), min(c_1c), min(c_time), min(c_timetz), min(c_timestamp), min(c_timestamptz), min(c_xid8), range_intersect_agg(c_int4range), range_intersect_agg(c_int4multirange), regr_count(((2 * b) + 3)::double precision, b::double precision), sum(b::real), sum(b::double precision), sum(b::smallint), sum(b), sum(c_interval), sum(c_money) FROM public.pagg_tab_p2 WHERE ((c_serial >= 1)) AND ((c_serial <= 30)) + -> Foreign Scan + Output: (PARTIAL array_agg(pagg_tab_2.c_int4array)), (PARTIAL array_agg(pagg_tab_2.b)), (PARTIAL avg((pagg_tab_2.b)::smallint)), (PARTIAL avg(pagg_tab_2.b)), (PARTIAL avg((pagg_tab_2.b)::bigint)), (PARTIAL avg(pagg_tab_2.c_interval)), (PARTIAL avg((pagg_tab_2.b)::real)), (PARTIAL avg((pagg_tab_2.b)::double precision)), (PARTIAL avg((pagg_tab_2.b)::numeric)), (PARTIAL corr((pagg_tab_2.b)::double precision, ((pagg_tab_2.b * pagg_tab_2.b))::double precision)), (PARTIAL covar_pop((pagg_tab_2.b)::double precision, ((pagg_tab_2.b * pagg_tab_2.b))::double precision)), (PARTIAL covar_samp((pagg_tab_2.b)::double precision, ((pagg_tab_2.b * pagg_tab_2.b))::double precision)), (PARTIAL regr_avgx(((2 * pagg_tab_2.b))::double precision, (pagg_tab_2.b)::double precision)), (PARTIAL regr_avgy(((2 * pagg_tab_2.b))::double precision, (pagg_tab_2.b)::double precision)), (PARTIAL regr_intercept((((2 * pagg_tab_2.b) + 3))::double precision, (pagg_tab_2.b)::double precision)), (PARTIAL regr_r2(((pagg_tab_2.b * pagg_tab_2.b))::double precision, (pagg_tab_2.b)::double precision)), (PARTIAL regr_slope((((2 * pagg_tab_2.b) + 3))::double precision, (pagg_tab_2.b)::double precision)), (PARTIAL regr_sxx((((2 * pagg_tab_2.b) + 3))::double precision, (pagg_tab_2.b)::double precision)), (PARTIAL regr_sxy(((pagg_tab_2.b * pagg_tab_2.b))::double precision, (pagg_tab_2.b)::double precision)), (PARTIAL regr_syy((((2 * pagg_tab_2.b) + 3))::double precision, (pagg_tab_2.b)::double precision)), (PARTIAL stddev((pagg_tab_2.b)::real)), (PARTIAL stddev((pagg_tab_2.b)::double precision)), (PARTIAL stddev((pagg_tab_2.b)::smallint)), (PARTIAL stddev(pagg_tab_2.b)), (PARTIAL stddev((pagg_tab_2.b)::bigint)), (PARTIAL stddev((pagg_tab_2.b)::numeric)), (PARTIAL stddev_pop((pagg_tab_2.b)::real)), (PARTIAL stddev_pop((pagg_tab_2.b)::double precision)), (PARTIAL stddev_pop((pagg_tab_2.b)::smallint)), (PARTIAL stddev_pop(pagg_tab_2.b)), (PARTIAL stddev_pop((pagg_tab_2.b)::bigint)), (PARTIAL stddev_pop((pagg_tab_2.b)::numeric)), (PARTIAL stddev_samp((pagg_tab_2.b)::real)), (PARTIAL stddev_samp((pagg_tab_2.b)::double precision)), (PARTIAL stddev_samp((pagg_tab_2.b)::smallint)), (PARTIAL stddev_samp(pagg_tab_2.b)), (PARTIAL stddev_samp((pagg_tab_2.b)::bigint)), (PARTIAL stddev_samp((pagg_tab_2.b)::numeric)), (PARTIAL string_agg(pagg_tab_2.c_1c, ','::text)), (PARTIAL string_agg(pagg_tab_2.c_1b, '\x2c'::bytea)), (PARTIAL sum((pagg_tab_2.b)::bigint)), (PARTIAL sum((pagg_tab_2.b)::numeric)), (PARTIAL variance((pagg_tab_2.b)::real)), (PARTIAL variance((pagg_tab_2.b)::double precision)), (PARTIAL variance((pagg_tab_2.b)::smallint)), (PARTIAL variance(pagg_tab_2.b)), (PARTIAL variance((pagg_tab_2.b)::bigint)), (PARTIAL variance((pagg_tab_2.b)::numeric)), (PARTIAL var_pop((pagg_tab_2.b)::real)), (PARTIAL var_pop((pagg_tab_2.b)::double precision)), (PARTIAL var_pop((pagg_tab_2.b)::smallint)), (PARTIAL var_pop(pagg_tab_2.b)), (PARTIAL var_pop((pagg_tab_2.b)::bigint)), (PARTIAL var_pop((pagg_tab_2.b)::numeric)), (PARTIAL var_samp((pagg_tab_2.b)::real)), (PARTIAL var_samp((pagg_tab_2.b)::double precision)), (PARTIAL var_samp((pagg_tab_2.b)::smallint)), (PARTIAL var_samp(pagg_tab_2.b)), (PARTIAL var_samp((pagg_tab_2.b)::bigint)), (PARTIAL var_samp((pagg_tab_2.b)::numeric)), (PARTIAL any_value((pagg_tab_2.b * 0))), (PARTIAL bit_and(pagg_tab_2.c_bit)), (PARTIAL bit_and(pagg_tab_2.c_1or3int2)), (PARTIAL bit_and(pagg_tab_2.c_1or3int4)), (PARTIAL bit_and(pagg_tab_2.c_1or3int8)), (PARTIAL bit_or(pagg_tab_2.c_bit)), (PARTIAL bit_or(pagg_tab_2.c_1or3int2)), (PARTIAL bit_or(pagg_tab_2.c_1or3int4)), (PARTIAL bit_or(pagg_tab_2.c_1or3int8)), (PARTIAL bit_xor(pagg_tab_2.c_bit)), (PARTIAL bit_xor(pagg_tab_2.c_1or3int2)), (PARTIAL bit_xor(pagg_tab_2.c_1or3int4)), (PARTIAL bit_xor(pagg_tab_2.c_1or3int8)), (PARTIAL bool_and(pagg_tab_2.c_bool)), (PARTIAL bool_or(pagg_tab_2.c_bool)), (PARTIAL count(pagg_tab_2.b)), (PARTIAL count(*)), (PARTIAL every(pagg_tab_2.c_bool)), (PARTIAL max(pagg_tab_2.c_int4array)), (PARTIAL max(pagg_tab_2.c_enum)), (PARTIAL max((pagg_tab_2.c_1c)::character(1))), (PARTIAL max(('01-01-2000'::date + pagg_tab_2.b))), (PARTIAL max(('0.0.0.0'::inet + (pagg_tab_2.b)::bigint))), (PARTIAL max((pagg_tab_2.b)::real)), (PARTIAL max((pagg_tab_2.b)::double precision)), (PARTIAL max((pagg_tab_2.b)::smallint)), (PARTIAL max(pagg_tab_2.b)), (PARTIAL max((pagg_tab_2.b)::bigint)), (PARTIAL max(pagg_tab_2.c_interval)), (PARTIAL max(pagg_tab_2.c_money)), (PARTIAL max((pagg_tab_2.b)::numeric)), (PARTIAL max((pagg_tab_2.b)::oid)), (PARTIAL max(pagg_tab_2.c_pg_lsn)), (PARTIAL max(pagg_tab_2.c_tid)), (PARTIAL max(pagg_tab_2.c_1c)), (PARTIAL max(pagg_tab_2.c_time)), (PARTIAL max(pagg_tab_2.c_timetz)), (PARTIAL max(pagg_tab_2.c_timestamp)), (PARTIAL max(pagg_tab_2.c_timestamptz)), (PARTIAL max(pagg_tab_2.c_xid8)), (PARTIAL min(pagg_tab_2.c_int4array)), (PARTIAL min(pagg_tab_2.c_enum)), (PARTIAL min((pagg_tab_2.c_1c)::character(1))), (PARTIAL min(('01-01-2000'::date + pagg_tab_2.b))), (PARTIAL min(('0.0.0.0'::inet + (pagg_tab_2.b)::bigint))), (PARTIAL min((pagg_tab_2.b)::real)), (PARTIAL min((pagg_tab_2.b)::double precision)), (PARTIAL min((pagg_tab_2.b)::smallint)), (PARTIAL min(pagg_tab_2.b)), (PARTIAL min((pagg_tab_2.b)::bigint)), (PARTIAL min(pagg_tab_2.c_interval)), (PARTIAL min(pagg_tab_2.c_money)), (PARTIAL min((pagg_tab_2.b)::numeric)), (PARTIAL min((pagg_tab_2.b)::oid)), (PARTIAL min(pagg_tab_2.c_pg_lsn)), (PARTIAL min(pagg_tab_2.c_tid)), (PARTIAL min(pagg_tab_2.c_1c)), (PARTIAL min(pagg_tab_2.c_time)), (PARTIAL min(pagg_tab_2.c_timetz)), (PARTIAL min(pagg_tab_2.c_timestamp)), (PARTIAL min(pagg_tab_2.c_timestamptz)), (PARTIAL min(pagg_tab_2.c_xid8)), (PARTIAL range_intersect_agg(pagg_tab_2.c_int4range)), (PARTIAL range_intersect_agg(pagg_tab_2.c_int4multirange)), (PARTIAL regr_count((((2 * pagg_tab_2.b) + 3))::double precision, (pagg_tab_2.b)::double precision)), (PARTIAL sum((pagg_tab_2.b)::real)), (PARTIAL sum((pagg_tab_2.b)::double precision)), (PARTIAL sum((pagg_tab_2.b)::smallint)), (PARTIAL sum(pagg_tab_2.b)), (PARTIAL sum(pagg_tab_2.c_interval)), (PARTIAL sum(pagg_tab_2.c_money)) + Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2) + Remote SQL: SELECT array_agg_p_anyarray(c_int4array), array_agg_p_anynonarray(b), avg_p_int2(b::smallint), avg_p_int4(b), avg_p_int8(b::bigint), avg_p_interval(c_interval), avg_p_float4(b::real), avg_p_float8(b::double precision), avg_p_numeric(b::numeric), corr_p(b::double precision, (b * b)::double precision), covar_pop_p(b::double precision, (b * b)::double precision), covar_samp_p(b::double precision, (b * b)::double precision), regr_avgx_p((2 * b)::double precision, b::double precision), regr_avgy_p((2 * b)::double precision, b::double precision), regr_intercept_p(((2 * b) + 3)::double precision, b::double precision), regr_r2_p((b * b)::double precision, b::double precision), regr_slope_p(((2 * b) + 3)::double precision, b::double precision), regr_sxx_p(((2 * b) + 3)::double precision, b::double precision), regr_sxy_p((b * b)::double precision, b::double precision), regr_syy_p(((2 * b) + 3)::double precision, b::double precision), stddev_p_float4(b::real), stddev_p_float8(b::double precision), stddev_p_int2(b::smallint), stddev_p_int4(b), stddev_p_int8(b::bigint), stddev_p_numeric(b::numeric), stddev_pop_p_float4(b::real), stddev_pop_p_float8(b::double precision), stddev_pop_p_int2(b::smallint), stddev_pop_p_int4(b), stddev_pop_p_int8(b::bigint), stddev_pop_p_numeric(b::numeric), stddev_samp_p_float4(b::real), stddev_samp_p_float8(b::double precision), stddev_samp_p_int2(b::smallint), stddev_samp_p_int4(b), stddev_samp_p_int8(b::bigint), stddev_samp_p_numeric(b::numeric), string_agg_p_text_text(c_1c, ','::text), string_agg_p_bytea_bytea(c_1b, E'\\x2c'::bytea), sum_p_int8(b::bigint), sum_p_numeric(b::numeric), variance_p_float4(b::real), variance_p_float8(b::double precision), variance_p_int2(b::smallint), variance_p_int4(b), variance_p_int8(b::bigint), variance_p_numeric(b::numeric), var_pop_p_float4(b::real), var_pop_p_float8(b::double precision), var_pop_p_int2(b::smallint), var_pop_p_int4(b), var_pop_p_int8(b::bigint), var_pop_p_numeric(b::numeric), var_samp_p_float4(b::real), var_samp_p_float8(b::double precision), var_samp_p_int2(b::smallint), var_samp_p_int4(b), var_samp_p_int8(b::bigint), var_samp_p_numeric(b::numeric), any_value((b * 0)), bit_and(c_bit), bit_and(c_1or3int2), bit_and(c_1or3int4), bit_and(c_1or3int8), bit_or(c_bit), bit_or(c_1or3int2), bit_or(c_1or3int4), bit_or(c_1or3int8), bit_xor(c_bit), bit_xor(c_1or3int2), bit_xor(c_1or3int4), bit_xor(c_1or3int8), bool_and(c_bool), bool_or(c_bool), count(b), count(*), every(c_bool), max(c_int4array), max(c_enum), max(c_1c::character(1)), max(('01-01-2000'::date + b)), max(('0.0.0.0'::inet + b)), max(b::real), max(b::double precision), max(b::smallint), max(b), max(b::bigint), max(c_interval), max(c_money), max(b::numeric), max(b::oid), max(c_pg_lsn), max(c_tid), max(c_1c), max(c_time), max(c_timetz), max(c_timestamp), max(c_timestamptz), max(c_xid8), min(c_int4array), min(c_enum), min(c_1c::character(1)), min(('01-01-2000'::date + b)), min(('0.0.0.0'::inet + b)), min(b::real), min(b::double precision), min(b::smallint), min(b), min(b::bigint), min(c_interval), min(c_money), min(b::numeric), min(b::oid), min(c_pg_lsn), min(c_tid), min(c_1c), min(c_time), min(c_timetz), min(c_timestamp), min(c_timestamptz), min(c_xid8), range_intersect_agg(c_int4range), range_intersect_agg(c_int4multirange), regr_count(((2 * b) + 3)::double precision, b::double precision), sum(b::real), sum(b::double precision), sum(b::smallint), sum(b), sum(c_interval), sum(c_money) FROM public.pagg_tab_p3 WHERE ((c_serial >= 1)) AND ((c_serial <= 30)) +(15 rows) + +SELECT /* aggregate function <> aggpartialfunc */ + array_agg(c_int4array), array_agg(b), + avg(b::int2), avg(b::int4), avg(b::int8), avg(c_interval), + avg(b::float4), avg(b::float8), + corr(b::float8, (b * b)::float8), + covar_pop(b::float8, (b * b)::float8), + covar_samp(b::float8, (b * b)::float8), + regr_avgx((2 * b)::float8, b::float8), + regr_avgy((2 * b)::float8, b::float8), + regr_intercept((2 * b + 3)::float8, b::float8), + regr_r2((b * b)::float8, b::float8), + regr_slope((2 * b + 3)::float8, b::float8), + regr_sxx((2 * b + 3)::float8, b::float8), + regr_sxy((b * b)::float8, b::float8), + regr_syy((2 * b + 3)::float8, b::float8), + stddev(b::float4), stddev(b::float8), stddev(b::int2), stddev(b::int4), stddev(b::int8), stddev(b::numeric), + stddev_pop(b::float4), stddev_pop(b::float8), stddev_pop(b::int2), stddev_pop(b::int4), stddev_pop(b::int8), stddev_pop(b::numeric), + stddev_samp(b::float4), stddev_samp(b::float8), stddev_samp(b::int2), stddev_samp(b::int4), stddev_samp(b::int8), stddev_samp(b::numeric), + string_agg(c_1c, ','), string_agg(c_1b, ','), + sum(b::int8), sum(b::numeric), + variance(b::float4), variance(b::float8), variance(b::int2), variance(b::int4), variance(b::int8), variance(b::numeric), + var_pop(b::float4), var_pop(b::float8), var_pop(b::int2), var_pop(b::int4), var_pop(b::int8), var_pop(b::numeric), + var_samp(b::float4), var_samp(b::float8), var_samp(b::int2), var_samp(b::int4), var_samp(b::int8), var_samp(b::numeric), + /* aggregate function = aggpartialfunc */ + any_value(b * 0), + bit_and(c_bit), bit_and(c_1or3int2), bit_and(c_1or3int4), bit_and(c_1or3int8), + bit_or(c_bit), bit_or(c_1or3int2), bit_or(c_1or3int4), bit_or(c_1or3int8), + bit_xor(c_bit), bit_xor(c_1or3int2), bit_xor(c_1or3int4), bit_xor(c_1or3int8), + bool_and(c_bool), + bool_or(c_bool), + count(b), count(*), + every(c_bool), + max(c_int4array), max(c_enum), max(c_1c::char(1)), max('2000-01-01'::date + b), max('0.0.0.0'::inet + b), max(b::float4), max(b::float8), max(b::int2), max(b::int4), max(b::int8), max(c_interval), max(c_money), max(b::numeric), max(b::oid), max(c_pg_lsn), max(c_tid), max(c_1c), max(c_time), max(c_timetz), max(c_timestamp), max(c_timestamptz), max(c_xid8), + min(c_int4array), min(c_enum), min(c_1c::char(1)), min('2000-01-01'::date + b), min('0.0.0.0'::inet + b), min(b::float4), min(b::float8), min(b::int2), min(b::int4), min(b::int8), min(c_interval), min(c_money), min(b::numeric), min(b::oid), min(c_pg_lsn), min(c_tid), min(c_1c), min(c_time), min(c_timetz), min(c_timestamp), min(c_timestamptz), min(c_xid8), + range_intersect_agg(c_int4range), range_intersect_agg(c_int4multirange), + regr_count((2 * b + 3)::float8, b::float8), + sum(b::float4), sum(b::float8), sum(b::int2), sum(b::int4), sum(c_interval), sum(c_money) + FROM pagg_tab WHERE c_serial between 1 and 30; + array_agg | array_agg | avg | avg | avg | avg | avg | avg | corr | covar_pop | covar_samp | regr_avgx | regr_avgy | regr_intercept | regr_r2 | regr_slope | regr_sxx | regr_sxy | regr_syy | stddev | stddev | stddev | stddev | stddev | stddev | stddev_pop | stddev_pop | stddev_pop | stddev_pop | stddev_pop | stddev_pop | stddev_samp | stddev_samp | stddev_samp | stddev_samp | stddev_samp | stddev_samp | string_agg | string_agg | sum | sum | variance | variance | variance | variance | variance | variance | var_pop | var_pop | var_pop | var_pop | var_pop | var_pop | var_samp | var_samp | var_samp | var_samp | var_samp | var_samp | any_value | bit_and | bit_and | bit_and | bit_and | bit_or | bit_or | bit_or | bit_or | bit_xor | bit_xor | bit_xor | bit_xor | bool_and | bool_or | count | count | every | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | min | min | min | min | min | min | min | min | min | min | min | min | min | min | min | min | min | min | min | min | min | min | range_intersect_agg | range_intersect_agg | regr_count | sum | sum | sum | sum | sum | sum +---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+------------------------------------------------------------------------------------+---------------------+---------------------+---------------------+------------+------+------+--------------------+--------------------+------------+-----------+-----------+----------------+--------------------+------------+----------+----------+----------+-------------------+-------------------+--------------------+--------------------+--------------------+--------------------+------------------+------------------+--------------------+--------------------+--------------------+--------------------+-------------------+-------------------+--------------------+--------------------+--------------------+--------------------+-------------------------------------------------------------+--------------------------------------------------------------------------------------------------------------------------+-----+-----+----------+----------+---------------------+---------------------+---------------------+---------------------+-------------------+-------------------+---------------------+---------------------+---------------------+---------------------+----------+----------+---------------------+---------------------+---------------------+---------------------+-----------+---------+---------+---------+---------+--------+--------+--------+--------+---------+---------+---------+---------+----------+---------+-------+-------+-------+-------+-------+-----+------------+----------+-----+-----+-----+-----+-----+---------+-------+-----+-----+------+--------+-----+----------+-------------+--------------------------+------------------------------+-----+-------+-----+-----+------------+---------+-----+-----+-----+-----+-----+-----+-------+-----+-----+-----+-------+-----+----------+-------------+--------------------------+------------------------------+-----+---------------------+---------------------+------------+-----+-----+-----+-----+-----------+-------- + {{1,0},{0,0},{1,0},{0,0},{1,0},{0,0},{1,0},{0,0},{1,0},{0,0},{0,0},{1,0},{0,0},{1,0},{0,0},{1,0},{0,0},{1,0},{0,0},{1,0},{0,0},{1,0},{0,0},{1,0},{0,0},{1,0},{0,0},{1,0},{0,0},{1,0}} | {1,2,3,4,5,6,7,8,9,30,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29} | 15.5000000000000000 | 15.5000000000000000 | 15.5000000000000000 | @ 0.5 secs | 15.5 | 15.5 | 0.9702989135892578 | 2322.4166666666665 | 2402.5 | 15.5 | 31 | 3 | 0.9414799817124941 | 2 | 2247.5 | 69672.5 | 8990 | 8.803408430829505 | 8.803408430829505 | 8.8034084308295046 | 8.8034084308295046 | 8.8034084308295046 | 8.8034084308295046 | 8.65544144839919 | 8.65544144839919 | 8.6554414483991899 | 8.6554414483991899 | 8.6554414483991899 | 8.6554414483991899 | 8.803408430829505 | 8.803408430829505 | 8.8034084308295046 | 8.8034084308295046 | 8.8034084308295046 | 8.8034084308295046 | 0,1,2,3,4,5,6,7,8,9,9,0,1,2,3,4,5,6,7,8,9,0,1,2,3,4,5,6,7,8 | \x302c312c322c332c342c352c362c372c382c392c392c302c312c322c332c342c352c362c372c382c392c302c312c322c332c342c352c362c372c38 | 465 | 465 | 77.5 | 77.5 | 77.5000000000000000 | 77.5000000000000000 | 77.5000000000000000 | 77.5000000000000000 | 74.91666666666667 | 74.91666666666667 | 74.9166666666666667 | 74.9166666666666667 | 74.9166666666666667 | 74.9166666666666667 | 77.5 | 77.5 | 77.5000000000000000 | 77.5000000000000000 | 77.5000000000000000 | 77.5000000000000000 | 0 | 01 | 1 | 1 | 1 | 11 | 3 | 3 | 3 | 10 | 2 | 2 | 2 | f | t | 30 | 30 | f | {1,0} | happy | 9 | 01-31-2000 | 0.0.0.30 | 30 | 30 | 30 | 30 | 30 | @ 1 sec | $1.00 | 30 | 30 | 0/30 | (0,30) | 9 | 00:00:30 | 00:00:30-07 | Sat Jan 01 00:00:30 2000 | Sat Jan 01 00:00:30 2000 PST | 9 | {0,0} | sad | 0 | 01-02-2000 | 0.0.0.1 | 1 | 1 | 1 | 1 | 1 | @ 0 | $0.00 | 1 | 1 | 0/1 | (0,1) | 0 | 00:00:01 | 00:00:01-07 | Sat Jan 01 00:00:01 2000 | Sat Jan 01 00:00:01 2000 PST | 0 | [0,1) | {[0,1),[100,101)} | 30 | 465 | 465 | 465 | 465 | @ 15 secs | $15.00 +(1 row) + +-- Tests for backward compatibility +ALTER SERVER loopback OPTIONS (ADD check_partial_aggregate_support 'false'); +EXPLAIN (VERBOSE, COSTS OFF) +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; + QUERY PLAN +------------------------------------------------------------------------------------------------------------------------ + Sort + Output: pagg_tab.b, (avg(pagg_tab.a)), (max(pagg_tab.a)), (count(*)) + Sort Key: pagg_tab.b + -> Finalize HashAggregate + Output: pagg_tab.b, avg(pagg_tab.a), max(pagg_tab.a), count(*) + Group Key: pagg_tab.b + -> Append + -> Foreign Scan + Output: pagg_tab.b, (PARTIAL avg(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab) + Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p1 GROUP BY 1 + -> Foreign Scan + Output: pagg_tab_1.b, (PARTIAL avg(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1) + Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p2 GROUP BY 1 + -> Foreign Scan + Output: pagg_tab_2.b, (PARTIAL avg(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2) + Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p3 GROUP BY 1 +(19 rows) + +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; + b | avg | max | count +----+---------------------+-----+------- + 0 | 10.0000000000000000 | 20 | 60 + 1 | 11.0000000000000000 | 21 | 60 + 2 | 12.0000000000000000 | 22 | 60 + 3 | 13.0000000000000000 | 23 | 60 + 4 | 14.0000000000000000 | 24 | 60 + 5 | 15.0000000000000000 | 25 | 60 + 6 | 16.0000000000000000 | 26 | 60 + 7 | 17.0000000000000000 | 27 | 60 + 8 | 18.0000000000000000 | 28 | 60 + 9 | 19.0000000000000000 | 29 | 60 + 10 | 10.0000000000000000 | 20 | 60 + 11 | 11.0000000000000000 | 21 | 60 + 12 | 12.0000000000000000 | 22 | 60 + 13 | 13.0000000000000000 | 23 | 60 + 14 | 14.0000000000000000 | 24 | 60 + 15 | 15.0000000000000000 | 25 | 60 + 16 | 16.0000000000000000 | 26 | 60 + 17 | 17.0000000000000000 | 27 | 60 + 18 | 18.0000000000000000 | 28 | 60 + 19 | 19.0000000000000000 | 29 | 60 + 20 | 10.0000000000000000 | 20 | 60 + 21 | 11.0000000000000000 | 21 | 60 + 22 | 12.0000000000000000 | 22 | 60 + 23 | 13.0000000000000000 | 23 | 60 + 24 | 14.0000000000000000 | 24 | 60 + 25 | 15.0000000000000000 | 25 | 60 + 26 | 16.0000000000000000 | 26 | 60 + 27 | 17.0000000000000000 | 27 | 60 + 28 | 18.0000000000000000 | 28 | 60 + 29 | 19.0000000000000000 | 29 | 60 + 30 | 10.0000000000000000 | 20 | 60 + 31 | 11.0000000000000000 | 21 | 60 + 32 | 12.0000000000000000 | 22 | 60 + 33 | 13.0000000000000000 | 23 | 60 + 34 | 14.0000000000000000 | 24 | 60 + 35 | 15.0000000000000000 | 25 | 60 + 36 | 16.0000000000000000 | 26 | 60 + 37 | 17.0000000000000000 | 27 | 60 + 38 | 18.0000000000000000 | 28 | 60 + 39 | 19.0000000000000000 | 29 | 60 + 40 | 10.0000000000000000 | 20 | 60 + 41 | 11.0000000000000000 | 21 | 60 + 42 | 12.0000000000000000 | 22 | 60 + 43 | 13.0000000000000000 | 23 | 60 + 44 | 14.0000000000000000 | 24 | 60 + 45 | 15.0000000000000000 | 25 | 60 + 46 | 16.0000000000000000 | 26 | 60 + 47 | 17.0000000000000000 | 27 | 60 + 48 | 18.0000000000000000 | 28 | 60 + 49 | 19.0000000000000000 | 29 | 60 +(50 rows) + +ALTER SERVER loopback OPTIONS (SET check_partial_aggregate_support 'true'); +EXPLAIN (VERBOSE, COSTS OFF) +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; + QUERY PLAN +------------------------------------------------------------------------------------------------------------------------ + Sort + Output: pagg_tab.b, (avg(pagg_tab.a)), (max(pagg_tab.a)), (count(*)) + Sort Key: pagg_tab.b + -> Finalize HashAggregate + Output: pagg_tab.b, avg(pagg_tab.a), max(pagg_tab.a), count(*) + Group Key: pagg_tab.b + -> Append + -> Foreign Scan + Output: pagg_tab.b, (PARTIAL avg(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab) + Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p1 GROUP BY 1 + -> Foreign Scan + Output: pagg_tab_1.b, (PARTIAL avg(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1) + Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p2 GROUP BY 1 + -> Foreign Scan + Output: pagg_tab_2.b, (PARTIAL avg(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2) + Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p3 GROUP BY 1 +(19 rows) + +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; + b | avg | max | count +----+---------------------+-----+------- + 0 | 10.0000000000000000 | 20 | 60 + 1 | 11.0000000000000000 | 21 | 60 + 2 | 12.0000000000000000 | 22 | 60 + 3 | 13.0000000000000000 | 23 | 60 + 4 | 14.0000000000000000 | 24 | 60 + 5 | 15.0000000000000000 | 25 | 60 + 6 | 16.0000000000000000 | 26 | 60 + 7 | 17.0000000000000000 | 27 | 60 + 8 | 18.0000000000000000 | 28 | 60 + 9 | 19.0000000000000000 | 29 | 60 + 10 | 10.0000000000000000 | 20 | 60 + 11 | 11.0000000000000000 | 21 | 60 + 12 | 12.0000000000000000 | 22 | 60 + 13 | 13.0000000000000000 | 23 | 60 + 14 | 14.0000000000000000 | 24 | 60 + 15 | 15.0000000000000000 | 25 | 60 + 16 | 16.0000000000000000 | 26 | 60 + 17 | 17.0000000000000000 | 27 | 60 + 18 | 18.0000000000000000 | 28 | 60 + 19 | 19.0000000000000000 | 29 | 60 + 20 | 10.0000000000000000 | 20 | 60 + 21 | 11.0000000000000000 | 21 | 60 + 22 | 12.0000000000000000 | 22 | 60 + 23 | 13.0000000000000000 | 23 | 60 + 24 | 14.0000000000000000 | 24 | 60 + 25 | 15.0000000000000000 | 25 | 60 + 26 | 16.0000000000000000 | 26 | 60 + 27 | 17.0000000000000000 | 27 | 60 + 28 | 18.0000000000000000 | 28 | 60 + 29 | 19.0000000000000000 | 29 | 60 + 30 | 10.0000000000000000 | 20 | 60 + 31 | 11.0000000000000000 | 21 | 60 + 32 | 12.0000000000000000 | 22 | 60 + 33 | 13.0000000000000000 | 23 | 60 + 34 | 14.0000000000000000 | 24 | 60 + 35 | 15.0000000000000000 | 25 | 60 + 36 | 16.0000000000000000 | 26 | 60 + 37 | 17.0000000000000000 | 27 | 60 + 38 | 18.0000000000000000 | 28 | 60 + 39 | 19.0000000000000000 | 29 | 60 + 40 | 10.0000000000000000 | 20 | 60 + 41 | 11.0000000000000000 | 21 | 60 + 42 | 12.0000000000000000 | 22 | 60 + 43 | 13.0000000000000000 | 23 | 60 + 44 | 14.0000000000000000 | 24 | 60 + 45 | 15.0000000000000000 | 25 | 60 + 46 | 16.0000000000000000 | 26 | 60 + 47 | 17.0000000000000000 | 27 | 60 + 48 | 18.0000000000000000 | 28 | 60 + 49 | 19.0000000000000000 | 29 | 60 +(50 rows) + +ALTER SERVER loopback OPTIONS (DROP check_partial_aggregate_support); +-- List of aggregate functions such that the aggregate function supports partial aggregate +-- and aggpartialfn field of the aggregate function is invalid +SELECT aggfnoid::text + FROM pg_aggregate a JOIN pg_type b ON a.aggtranstype = b.oid + WHERE (aggcombinefn::text <> '-') AND (aggpartialfn::text = '-') + ORDER BY 1; + aggfnoid +---------- +(0 rows) + +-- List of aggregate functions which have incorrect aggpartialfunc +SELECT a.aggfnoid + FROM pg_aggregate a JOIN pg_type b ON a.aggtranstype = b.oid + JOIN pg_aggregate c on a.aggpartialfn = c.aggfnoid + WHERE (a.aggcombinefn::text <> '-') + AND (a.aggpartialfn::text <> '-') + AND (a.aggtransfn <> c.aggtransfn + OR (c.aggcombinefn::text <> '-' + AND a.aggcombinefn <> c.aggcombinefn) + OR (a.agginitval IS NOT NULL + AND c.agginitval IS NOT NULL + AND a.agginitval <> c.agginitval) + OR (b.typname = 'internal' + AND c.aggserialfn::text <> '-' + AND a.aggserialfn <> c.aggserialfn) + OR (b.typname = 'internal' + AND c.aggdeserialfn::text <> '-' + AND a.aggdeserialfn <> c.aggdeserialfn) + OR (b.typname = 'internal' + AND a.aggserialfn <> c.aggfinalfn) + OR (b.typname <> 'internal' + AND c.aggfinalfn::text <> '-')) + ORDER BY 1; + aggfnoid +---------- +(0 rows) + +-- It's unsafe to push down partial aggregates which contain distinct clause +EXPLAIN (VERBOSE, COSTS OFF) +SELECT max(a), count(distinct b) FROM pagg_tab; + QUERY PLAN +----------------------------------------------------------------------------------------- + Aggregate + Output: max(pagg_tab.a), count(DISTINCT pagg_tab.b) + -> Merge Append + Sort Key: pagg_tab.b + -> Foreign Scan on public.fpagg_tab_p1 pagg_tab_1 + Output: pagg_tab_1.a, pagg_tab_1.b + Remote SQL: SELECT a, b FROM public.pagg_tab_p1 ORDER BY b ASC NULLS LAST + -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_2 + Output: pagg_tab_2.a, pagg_tab_2.b + Remote SQL: SELECT a, b FROM public.pagg_tab_p2 ORDER BY b ASC NULLS LAST + -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_3 + Output: pagg_tab_3.a, pagg_tab_3.b + Remote SQL: SELECT a, b FROM public.pagg_tab_p3 ORDER BY b ASC NULLS LAST +(13 rows) + +SELECT max(a), count(distinct b) FROM pagg_tab; + max | count +-----+------- + 29 | 50 +(1 row) + +-- It's unsafe to push down partial aggregates which contain order by clause +EXPLAIN (VERBOSE, COSTS OFF) SELECT array_agg(b order by b) FROM pagg_tab WHERE c_serial between 1 and 30; + QUERY PLAN +----------------------------------------------------------------------------------------------------------------- + Aggregate + Output: array_agg(pagg_tab.b ORDER BY pagg_tab.b) + -> Sort + Output: pagg_tab.b + Sort Key: pagg_tab.b + -> Append + -> Foreign Scan on public.fpagg_tab_p1 pagg_tab_1 + Output: pagg_tab_1.b + Remote SQL: SELECT b FROM public.pagg_tab_p1 WHERE ((c_serial >= 1)) AND ((c_serial <= 30)) + -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_2 + Output: pagg_tab_2.b + Remote SQL: SELECT b FROM public.pagg_tab_p2 WHERE ((c_serial >= 1)) AND ((c_serial <= 30)) + -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_3 + Output: pagg_tab_3.b + Remote SQL: SELECT b FROM public.pagg_tab_p3 WHERE ((c_serial >= 1)) AND ((c_serial <= 30)) +(15 rows) + +SELECT array_agg(b order by b) FROM pagg_tab WHERE c_serial between 1 and 30; + array_agg +------------------------------------------------------------------------------------ + {1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30} +(1 row) + +-- Create user-defined aggregates whose stype is internal +CREATE SCHEMA postgres_fdw_test; +CREATE AGGREGATE postgres_fdw_test.postgres_fdw_sum_p_int8(int8) ( + SFUNC = int8_avg_accum, + STYPE = internal, + COMBINEFUNC = int8_avg_combine, + FINALFUNC = int8_avg_serialize, + SERIALFUNC = int8_avg_serialize, + DESERIALFUNC = int8_avg_deserialize, + AGGPARTIALFUNC = postgres_fdw_test.postgres_fdw_sum_p_int8 +); +CREATE AGGREGATE postgres_fdw_sum(int8) ( + SFUNC = int8_avg_accum, + STYPE = internal, + COMBINEFUNC = int8_avg_combine, + FINALFUNC = numeric_poly_sum, + SERIALFUNC = int8_avg_serialize, + DESERIALFUNC = int8_avg_deserialize, + AGGPARTIALFUNC = postgres_fdw_test.postgres_fdw_sum_p_int8 +); +-- Create user-defined aggregates whose stype is not internal +CREATE AGGREGATE postgres_fdw_avg_p_int4(int4) ( + SFUNC = int4_avg_accum, + STYPE = _int8, + COMBINEFUNC = int4_avg_combine, + INITCOND = '{0,0}', + AGGPARTIALFUNC = postgres_fdw_avg_p_int4 +); +CREATE AGGREGATE postgres_fdw_avg(int4) ( + SFUNC = int4_avg_accum, + STYPE = _int8, + COMBINEFUNC = int4_avg_combine, + FINALFUNC = int8_avg, + INITCOND = '{0,0}', + AGGPARTIALFUNC = postgres_fdw_avg_p_int4 +); +CREATE AGGREGATE postgres_fdw_sum_noaggpartialfunc(int8) ( + SFUNC = int8_avg_accum, + STYPE = internal, + COMBINEFUNC = int8_avg_combine, + FINALFUNC = numeric_poly_sum, + SERIALFUNC = int8_avg_serialize, + DESERIALFUNC = int8_avg_deserialize +); +-- It's safe to push down partial aggregate for user-defined aggregate +-- when it belongs to an extension that's listed in extensions set +-- and when its aggpartialfunc belongs to an extension that's listed +-- in extensions set. +ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_sum(int8); +ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_test.postgres_fdw_sum_p_int8(int8); +ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_avg(int4); +ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_avg_p_int4(int4); +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_sum(b::int8) FROM pagg_tab; + QUERY PLAN +--------------------------------------------------------------------------------------------------------------- + Finalize Aggregate + Output: postgres_fdw_sum((pagg_tab.b)::bigint) + -> Append + -> Foreign Scan + Output: (PARTIAL postgres_fdw_sum((pagg_tab.b)::bigint)) + Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab) + Remote SQL: SELECT postgres_fdw_test.postgres_fdw_sum_p_int8(b::bigint) FROM public.pagg_tab_p1 + -> Foreign Scan + Output: (PARTIAL postgres_fdw_sum((pagg_tab_1.b)::bigint)) + Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1) + Remote SQL: SELECT postgres_fdw_test.postgres_fdw_sum_p_int8(b::bigint) FROM public.pagg_tab_p2 + -> Foreign Scan + Output: (PARTIAL postgres_fdw_sum((pagg_tab_2.b)::bigint)) + Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2) + Remote SQL: SELECT postgres_fdw_test.postgres_fdw_sum_p_int8(b::bigint) FROM public.pagg_tab_p3 +(15 rows) + +SELECT postgres_fdw_sum(b::int8) FROM pagg_tab; + postgres_fdw_sum +------------------ + 73500 +(1 row) + +-- It's unsafe to push down partial aggregate for user-defined aggregate +-- when it has no aggpartialfunc. +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_sum_noaggpartialfunc(b::int8) FROM pagg_tab; + QUERY PLAN +----------------------------------------------------------------------------------------- + Finalize Aggregate + Output: postgres_fdw_sum_noaggpartialfunc((pagg_tab.b)::bigint) + -> Append + -> Partial Aggregate + Output: PARTIAL postgres_fdw_sum_noaggpartialfunc((pagg_tab.b)::bigint) + -> Foreign Scan on public.fpagg_tab_p1 pagg_tab + Output: pagg_tab.b + Remote SQL: SELECT b FROM public.pagg_tab_p1 + -> Partial Aggregate + Output: PARTIAL postgres_fdw_sum_noaggpartialfunc((pagg_tab_1.b)::bigint) + -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1 + Output: pagg_tab_1.b + Remote SQL: SELECT b FROM public.pagg_tab_p2 + -> Partial Aggregate + Output: PARTIAL postgres_fdw_sum_noaggpartialfunc((pagg_tab_2.b)::bigint) + -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2 + Output: pagg_tab_2.b + Remote SQL: SELECT b FROM public.pagg_tab_p3 +(18 rows) + +SELECT postgres_fdw_sum_noaggpartialfunc(b::int8) FROM pagg_tab; + postgres_fdw_sum_noaggpartialfunc +----------------------------------- + 73500 +(1 row) + +ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_test.postgres_fdw_sum_p_int8(int8); +-- Reconnect for flushing the shippability cache +\c - +SET enable_partitionwise_aggregate TO true; +-- It's unsafe to push down partial aggregate for user-defined aggregate +-- when itself is aggpartialfunc and when it belongs to an extension +-- that's listed in extensions set and when stype is internal. +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_test.postgres_fdw_sum_p_int8(b::int8) FROM pagg_tab; + QUERY PLAN +------------------------------------------------------------------------------------------------- + Finalize Aggregate + Output: postgres_fdw_test.postgres_fdw_sum_p_int8((pagg_tab.b)::bigint) + -> Append + -> Partial Aggregate + Output: PARTIAL postgres_fdw_test.postgres_fdw_sum_p_int8((pagg_tab.b)::bigint) + -> Foreign Scan on public.fpagg_tab_p1 pagg_tab + Output: pagg_tab.b + Remote SQL: SELECT b FROM public.pagg_tab_p1 + -> Partial Aggregate + Output: PARTIAL postgres_fdw_test.postgres_fdw_sum_p_int8((pagg_tab_1.b)::bigint) + -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1 + Output: pagg_tab_1.b + Remote SQL: SELECT b FROM public.pagg_tab_p2 + -> Partial Aggregate + Output: PARTIAL postgres_fdw_test.postgres_fdw_sum_p_int8((pagg_tab_2.b)::bigint) + -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2 + Output: pagg_tab_2.b + Remote SQL: SELECT b FROM public.pagg_tab_p3 +(18 rows) + +SELECT postgres_fdw_test.postgres_fdw_sum_p_int8(b::int8) FROM pagg_tab; + postgres_fdw_sum_p_int8 +------------------------------------------------------------ + \x0000000000000bb80000000200000001000000000000000000070dac +(1 row) + +ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_sum(int8); +ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_test.postgres_fdw_sum_p_int8(int8); +-- Reconnect for flushing the shippability cache +\c - +SET enable_partitionwise_aggregate TO true; +-- It's unsafe to push down partial aggregate for user-defined aggregate +-- when it doesn't belong to an extension that's listed in extensions set +-- and when its aggpartialfunc belongs to an extension that's listed +-- in extensions set. +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_sum(b::int8) FROM pagg_tab; + QUERY PLAN +------------------------------------------------------------------------ + Finalize Aggregate + Output: postgres_fdw_sum((pagg_tab.b)::bigint) + -> Append + -> Partial Aggregate + Output: PARTIAL postgres_fdw_sum((pagg_tab.b)::bigint) + -> Foreign Scan on public.fpagg_tab_p1 pagg_tab + Output: pagg_tab.b + Remote SQL: SELECT b FROM public.pagg_tab_p1 + -> Partial Aggregate + Output: PARTIAL postgres_fdw_sum((pagg_tab_1.b)::bigint) + -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1 + Output: pagg_tab_1.b + Remote SQL: SELECT b FROM public.pagg_tab_p2 + -> Partial Aggregate + Output: PARTIAL postgres_fdw_sum((pagg_tab_2.b)::bigint) + -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2 + Output: pagg_tab_2.b + Remote SQL: SELECT b FROM public.pagg_tab_p3 +(18 rows) + +SELECT postgres_fdw_sum(b::int8) FROM pagg_tab; + postgres_fdw_sum +------------------ + 73500 +(1 row) + +-- It's safe to push down partial aggregate for user-defined aggregate +-- whose stype is not internal +-- when it belongs to an extension that's listed in extensions set +-- and when its aggpartialfunc belongs to an extension that's listed +-- in extensions set. +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_avg(b) FROM pagg_tab; + QUERY PLAN +-------------------------------------------------------------------------------------------- + Finalize Aggregate + Output: postgres_fdw_avg(pagg_tab.b) + -> Append + -> Foreign Scan + Output: (PARTIAL postgres_fdw_avg(pagg_tab.b)) + Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab) + Remote SQL: SELECT public.postgres_fdw_avg_p_int4(b) FROM public.pagg_tab_p1 + -> Foreign Scan + Output: (PARTIAL postgres_fdw_avg(pagg_tab_1.b)) + Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1) + Remote SQL: SELECT public.postgres_fdw_avg_p_int4(b) FROM public.pagg_tab_p2 + -> Foreign Scan + Output: (PARTIAL postgres_fdw_avg(pagg_tab_2.b)) + Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2) + Remote SQL: SELECT public.postgres_fdw_avg_p_int4(b) FROM public.pagg_tab_p3 +(15 rows) + +SELECT postgres_fdw_avg(b) FROM pagg_tab; + postgres_fdw_avg +--------------------- + 24.5000000000000000 +(1 row) + +-- It's safe to push down partial aggregate for user-defined aggregate +-- when itself is aggpartialfunc and when it belongs to an extension +-- that's listed in extensions set and when stype is not internal. +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_avg_p_int4(b) FROM pagg_tab; + QUERY PLAN +-------------------------------------------------------------------------------------------- + Finalize Aggregate + Output: postgres_fdw_avg_p_int4(pagg_tab.b) + -> Append + -> Foreign Scan + Output: (PARTIAL postgres_fdw_avg_p_int4(pagg_tab.b)) + Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab) + Remote SQL: SELECT public.postgres_fdw_avg_p_int4(b) FROM public.pagg_tab_p1 + -> Foreign Scan + Output: (PARTIAL postgres_fdw_avg_p_int4(pagg_tab_1.b)) + Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1) + Remote SQL: SELECT public.postgres_fdw_avg_p_int4(b) FROM public.pagg_tab_p2 + -> Foreign Scan + Output: (PARTIAL postgres_fdw_avg_p_int4(pagg_tab_2.b)) + Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2) + Remote SQL: SELECT public.postgres_fdw_avg_p_int4(b) FROM public.pagg_tab_p3 +(15 rows) + +SELECT postgres_fdw_avg_p_int4(b) FROM pagg_tab; + postgres_fdw_avg_p_int4 +------------------------- + {3000,73500} +(1 row) + +ALTER EXTENSION postgres_fdw DROP TYPE mood; +ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_test.postgres_fdw_sum_p_int8(int8); +ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_avg_p_int4(int4); +ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_avg(int4); +DROP AGGREGATE postgres_fdw_sum(int8); +DROP AGGREGATE postgres_fdw_test.postgres_fdw_sum_p_int8(int8); +DROP SCHEMA postgres_fdw_test; +DROP AGGREGATE postgres_fdw_avg(int4); +DROP AGGREGATE postgres_fdw_avg_p_int4(int4); +DROP AGGREGATE postgres_fdw_sum_noaggpartialfunc(int8); +ALTER SERVER loopback OPTIONS (DROP fdw_tuple_cost); -- =================================================================== -- access rights and superuser -- =================================================================== diff --git a/contrib/postgres_fdw/option.c b/contrib/postgres_fdw/option.c index 8c822f4ef9..67e15c8bb9 100644 --- a/contrib/postgres_fdw/option.c +++ b/contrib/postgres_fdw/option.c @@ -126,7 +126,8 @@ postgres_fdw_validator(PG_FUNCTION_ARGS) strcmp(def->defname, "async_capable") == 0 || strcmp(def->defname, "parallel_commit") == 0 || strcmp(def->defname, "parallel_abort") == 0 || - strcmp(def->defname, "keep_connections") == 0) + strcmp(def->defname, "keep_connections") == 0 || + strcmp(def->defname, "check_partial_aggregate_support") == 0) { /* these accept only boolean values */ (void) defGetBoolean(def); @@ -268,6 +269,7 @@ InitPgFdwOptions(void) /* batch_size is available on both server and table */ {"batch_size", ForeignServerRelationId, false}, {"batch_size", ForeignTableRelationId, false}, + {"check_partial_aggregate_support", ForeignServerRelationId, false}, /* async_capable is available on both server and table */ {"async_capable", ForeignServerRelationId, false}, {"async_capable", ForeignTableRelationId, false}, diff --git a/contrib/postgres_fdw/postgres_fdw.c b/contrib/postgres_fdw/postgres_fdw.c index c5cada55fb..5c48c5d50d 100644 --- a/contrib/postgres_fdw/postgres_fdw.c +++ b/contrib/postgres_fdw/postgres_fdw.c @@ -519,7 +519,7 @@ static bool foreign_join_ok(PlannerInfo *root, RelOptInfo *joinrel, JoinType jointype, RelOptInfo *outerrel, RelOptInfo *innerrel, JoinPathExtraData *extra); static bool foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel, - Node *havingQual); + Node *havingQual, PartitionwiseAggregateType patype); static List *get_useful_pathkeys_for_relation(PlannerInfo *root, RelOptInfo *rel); static List *get_useful_ecs_for_relation(PlannerInfo *root, RelOptInfo *rel); @@ -650,6 +650,8 @@ postgresGetForeignRelSize(PlannerInfo *root, fpinfo->fdw_tuple_cost = DEFAULT_FDW_TUPLE_COST; fpinfo->shippable_extensions = NIL; fpinfo->fetch_size = 100; + fpinfo->check_partial_aggregate_support = false; + fpinfo->remoteversion = 0; fpinfo->async_capable = false; apply_server_options(fpinfo); @@ -661,7 +663,7 @@ postgresGetForeignRelSize(PlannerInfo *root, * should match what ExecCheckPermissions() does. If we fail due to lack * of permissions, the query would have failed at runtime anyway. */ - if (fpinfo->use_remote_estimate) + if (fpinfo->use_remote_estimate || fpinfo->check_partial_aggregate_support) { Oid userid; @@ -6127,6 +6129,8 @@ apply_server_options(PgFdwRelationInfo *fpinfo) ExtractExtensionList(defGetString(def), false); else if (strcmp(def->defname, "fetch_size") == 0) (void) parse_int(defGetString(def), &fpinfo->fetch_size, 0, NULL); + else if (strcmp(def->defname, "check_partial_aggregate_support") == 0) + fpinfo->check_partial_aggregate_support = defGetBoolean(def); else if (strcmp(def->defname, "async_capable") == 0) fpinfo->async_capable = defGetBoolean(def); } @@ -6186,6 +6190,8 @@ merge_fdw_options(PgFdwRelationInfo *fpinfo, fpinfo->shippable_extensions = fpinfo_o->shippable_extensions; fpinfo->use_remote_estimate = fpinfo_o->use_remote_estimate; fpinfo->fetch_size = fpinfo_o->fetch_size; + fpinfo->check_partial_aggregate_support = fpinfo_o->check_partial_aggregate_support; + fpinfo->remoteversion = fpinfo_o->remoteversion; fpinfo->async_capable = fpinfo_o->async_capable; /* Merge the table level options from either side of the join. */ @@ -6366,7 +6372,7 @@ postgresGetForeignJoinPaths(PlannerInfo *root, */ static bool foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel, - Node *havingQual) + Node *havingQual, PartitionwiseAggregateType patype) { Query *query = root->parse; PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) grouped_rel->fdw_private; @@ -6380,6 +6386,10 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel, if (query->groupingSets) return false; + /* It's unsafe to push having statements with partial aggregates */ + if ((patype == PARTITIONWISE_AGGREGATE_PARTIAL) && havingQual) + return false; + /* Get the fpinfo of the underlying scan relation. */ ofpinfo = (PgFdwRelationInfo *) fpinfo->outerrel->fdw_private; @@ -6621,6 +6631,7 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage, /* Ignore stages we don't support; and skip any duplicate calls. */ if ((stage != UPPERREL_GROUP_AGG && + stage != UPPERREL_PARTIAL_GROUP_AGG && stage != UPPERREL_ORDERED && stage != UPPERREL_FINAL) || output_rel->fdw_private) @@ -6637,6 +6648,10 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage, add_foreign_grouping_paths(root, input_rel, output_rel, (GroupPathExtraData *) extra); break; + case UPPERREL_PARTIAL_GROUP_AGG: + add_foreign_grouping_paths(root, input_rel, output_rel, + (GroupPathExtraData *) extra); + break; case UPPERREL_ORDERED: add_foreign_ordered_paths(root, input_rel, output_rel); break; @@ -6677,7 +6692,8 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel, return; Assert(extra->patype == PARTITIONWISE_AGGREGATE_NONE || - extra->patype == PARTITIONWISE_AGGREGATE_FULL); + extra->patype == PARTITIONWISE_AGGREGATE_FULL || + extra->patype == PARTITIONWISE_AGGREGATE_PARTIAL); /* save the input_rel as outerrel in fpinfo */ fpinfo->outerrel = input_rel; @@ -6697,7 +6713,7 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel, * Use HAVING qual from extra. In case of child partition, it will have * translated Vars. */ - if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual)) + if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual, extra->patype)) return; /* diff --git a/contrib/postgres_fdw/postgres_fdw.h b/contrib/postgres_fdw/postgres_fdw.h index 02c1152319..96e73a10a8 100644 --- a/contrib/postgres_fdw/postgres_fdw.h +++ b/contrib/postgres_fdw/postgres_fdw.h @@ -86,6 +86,15 @@ typedef struct PgFdwRelationInfo ForeignServer *server; UserMapping *user; /* only set in use_remote_estimate mode */ + /* for partial aggregate pushdown */ + bool check_partial_aggregate_support; + + /* + * if remoteversion is zero, then that means the remote server version is + * unacquired. + */ + int remoteversion; + int fetch_size; /* fetch size for this remote table */ /* diff --git a/contrib/postgres_fdw/sql/postgres_fdw.sql b/contrib/postgres_fdw/sql/postgres_fdw.sql index b54903ad8f..827637b001 100644 --- a/contrib/postgres_fdw/sql/postgres_fdw.sql +++ b/contrib/postgres_fdw/sql/postgres_fdw.sql @@ -2963,16 +2963,31 @@ RESET enable_partitionwise_join; -- =================================================================== -- test partitionwise aggregates -- =================================================================== - -CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a); +ALTER SERVER loopback OPTIONS (ADD fdw_tuple_cost '0.5'); + +CREATE TYPE mood AS enum ('sad', 'ok', 'happy'); +ALTER EXTENSION postgres_fdw ADD TYPE mood; + +CREATE TABLE pagg_tab (a int, b int, c text, c_serial int4, + c_int4array _int4, c_interval interval, + c_money money, c_1c text, c_1b bytea, + c_bit bit(2), c_1or3int2 int2, + c_1or3int4 int4, c_1or3int8 int8, + c_bool bool, c_enum mood, c_pg_lsn pg_lsn, + c_tid tid, c_int4range int4range, + c_int4multirange int4multirange, + c_time time, c_timetz timetz, + c_timestamp timestamp, c_timestamptz timestamptz, + c_xid8 xid8) + PARTITION BY RANGE(a); CREATE TABLE pagg_tab_p1 (LIKE pagg_tab); CREATE TABLE pagg_tab_p2 (LIKE pagg_tab); CREATE TABLE pagg_tab_p3 (LIKE pagg_tab); -INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10; -INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10; -INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20; +INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i, array[(i % 2), 0], ((i % 2) || ' seconds')::interval, (i % 2)::money, ((i - 1) % 10)::text, (((i - 1) % 10)::text)::bytea, case when (i % 2) = 0 then B'01' else B'11' end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then true else false end, (case when (i % 2) = 0 then 'sad' else 'happy' end)::mood, ('0/' || i)::pg_lsn, ('(0,' || i || ')')::tid, int4range(0, i), int4multirange(int4range(0, i), int4range(100, 100 + i)), '00:00:00'::time + (i || ' seconds')::interval, '00:00:00'::timetz + (i || ' seconds')::interval, '2000-01-01'::timestamp + (i || ' seconds')::interval, '2000-01-01'::timestamptz + (i || ' seconds')::interval, ((i % 10)::text)::xid8 FROM generate_series(1, 3000) i WHERE (i % 30) < 10; +INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i, array[(i % 2), 0], ((i % 2) || ' seconds')::interval, (i % 2)::money, ((i - 1) % 10)::text, (((i - 1) % 10)::text)::bytea, case when (i % 2) = 0 then B'01' else B'11' end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then true else false end, (case when (i % 2) = 0 then 'sad' else 'happy' end)::mood, ('0/' || i)::pg_lsn, ('(0,' || i || ')')::tid, int4range(0, i), int4multirange(int4range(0, i), int4range(100, 100 + i)), '00:00:00'::time + (i || ' seconds')::interval, '00:00:00'::timetz + (i || ' seconds')::interval, '2000-01-01'::timestamp + (i || ' seconds')::interval, '2000-01-01'::timestamptz + (i || ' seconds')::interval, ((i % 10)::text)::xid8 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10; +INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i, array[(i % 2), 0], ((i % 2) || ' seconds')::interval, (i % 2)::money, ((i - 1) % 10)::text, (((i - 1) % 10)::text)::bytea, case when (i % 2) = 0 then B'01' else B'11' end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then true else false end, (case when (i % 2) = 0 then 'sad' else 'happy' end)::mood, ('0/' || i)::pg_lsn, ('(0,' || i || ')')::tid, int4range(0, i), int4multirange(int4range(0, i), int4range(100, 100 + i)), '00:00:00'::time + (i || ' seconds')::interval, '00:00:00'::timetz + (i || ' seconds')::interval, '2000-01-01'::timestamp + (i || ' seconds')::interval, '2000-01-01'::timestamptz + (i || ' seconds')::interval, ((i % 10)::text)::xid8 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20; -- Create foreign partitions CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1'); @@ -3006,6 +3021,275 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1; EXPLAIN (COSTS OFF) SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1; +-- It's unsafe to push down having clause when there are partial aggregates +EXPLAIN (VERBOSE, COSTS OFF) +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1; + +-- Partial aggregates are safe to push down without having clause +EXPLAIN (VERBOSE, COSTS OFF) +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; + +-- Partial aggregates are safe to push down for all built-in aggregates +EXPLAIN (VERBOSE, COSTS OFF) +SELECT /* aggregate function <> aggpartialfunc */ + array_agg(c_int4array), array_agg(b), + avg(b::int2), avg(b::int4), avg(b::int8), avg(c_interval), avg(b::float4), avg(b::float8), avg(b::numeric), + corr(b::float8, (b * b)::float8), + covar_pop(b::float8, (b * b)::float8), + covar_samp(b::float8, (b * b)::float8), + regr_avgx((2 * b)::float8, b::float8), + regr_avgy((2 * b)::float8, b::float8), + regr_intercept((2 * b + 3)::float8, b::float8), + regr_r2((b * b)::float8, b::float8), + regr_slope((2 * b + 3)::float8, b::float8), + regr_sxx((2 * b + 3)::float8, b::float8), + regr_sxy((b * b)::float8, b::float8), + regr_syy((2 * b + 3)::float8, b::float8), + stddev(b::float4), stddev(b::float8), stddev(b::int2), stddev(b::int4), stddev(b::int8), stddev(b::numeric), + stddev_pop(b::float4), stddev_pop(b::float8), stddev_pop(b::int2), stddev_pop(b::int4), stddev_pop(b::int8), stddev_pop(b::numeric), + stddev_samp(b::float4), stddev_samp(b::float8), stddev_samp(b::int2), stddev_samp(b::int4), stddev_samp(b::int8), stddev_samp(b::numeric), + string_agg(c_1c, ','), string_agg(c_1b, ','), + sum(b::int8), sum(b::numeric), + variance(b::float4), variance(b::float8), variance(b::int2), variance(b::int4), variance(b::int8), variance(b::numeric), + var_pop(b::float4), var_pop(b::float8), var_pop(b::int2), var_pop(b::int4), var_pop(b::int8), var_pop(b::numeric), + var_samp(b::float4), var_samp(b::float8), var_samp(b::int2), var_samp(b::int4), var_samp(b::int8), var_samp(b::numeric), + /* aggregate function = aggpartialfunc */ + any_value(b * 0), + bit_and(c_bit), bit_and(c_1or3int2), bit_and(c_1or3int4), bit_and(c_1or3int8), + bit_or(c_bit), bit_or(c_1or3int2), bit_or(c_1or3int4), bit_or(c_1or3int8), + bit_xor(c_bit), bit_xor(c_1or3int2), bit_xor(c_1or3int4), bit_xor(c_1or3int8), + bool_and(c_bool), + bool_or(c_bool), + count(b), count(*), + every(c_bool), + max(c_int4array), max(c_enum), max(c_1c::char(1)), max('2000-01-01'::date + b), max('0.0.0.0'::inet + b), max(b::float4), max(b::float8), max(b::int2), max(b::int4), max(b::int8), max(c_interval), max(c_money), max(b::numeric), max(b::oid), max(c_pg_lsn), max(c_tid), max(c_1c), max(c_time), max(c_timetz), max(c_timestamp), max(c_timestamptz), max(c_xid8), + min(c_int4array), min(c_enum), min(c_1c::char(1)), min('2000-01-01'::date + b), min('0.0.0.0'::inet + b), min(b::float4), min(b::float8), min(b::int2), min(b::int4), min(b::int8), min(c_interval), min(c_money), min(b::numeric), min(b::oid), min(c_pg_lsn), min(c_tid), min(c_1c), min(c_time), min(c_timetz), min(c_timestamp), min(c_timestamptz), min(c_xid8), + range_intersect_agg(c_int4range), range_intersect_agg(c_int4multirange), + regr_count((2 * b + 3)::float8, b::float8), + sum(b::float4), sum(b::float8), sum(b::int2), sum(b::int4), sum(c_interval), sum(c_money) + FROM pagg_tab WHERE c_serial between 1 and 30; + +SELECT /* aggregate function <> aggpartialfunc */ + array_agg(c_int4array), array_agg(b), + avg(b::int2), avg(b::int4), avg(b::int8), avg(c_interval), + avg(b::float4), avg(b::float8), + corr(b::float8, (b * b)::float8), + covar_pop(b::float8, (b * b)::float8), + covar_samp(b::float8, (b * b)::float8), + regr_avgx((2 * b)::float8, b::float8), + regr_avgy((2 * b)::float8, b::float8), + regr_intercept((2 * b + 3)::float8, b::float8), + regr_r2((b * b)::float8, b::float8), + regr_slope((2 * b + 3)::float8, b::float8), + regr_sxx((2 * b + 3)::float8, b::float8), + regr_sxy((b * b)::float8, b::float8), + regr_syy((2 * b + 3)::float8, b::float8), + stddev(b::float4), stddev(b::float8), stddev(b::int2), stddev(b::int4), stddev(b::int8), stddev(b::numeric), + stddev_pop(b::float4), stddev_pop(b::float8), stddev_pop(b::int2), stddev_pop(b::int4), stddev_pop(b::int8), stddev_pop(b::numeric), + stddev_samp(b::float4), stddev_samp(b::float8), stddev_samp(b::int2), stddev_samp(b::int4), stddev_samp(b::int8), stddev_samp(b::numeric), + string_agg(c_1c, ','), string_agg(c_1b, ','), + sum(b::int8), sum(b::numeric), + variance(b::float4), variance(b::float8), variance(b::int2), variance(b::int4), variance(b::int8), variance(b::numeric), + var_pop(b::float4), var_pop(b::float8), var_pop(b::int2), var_pop(b::int4), var_pop(b::int8), var_pop(b::numeric), + var_samp(b::float4), var_samp(b::float8), var_samp(b::int2), var_samp(b::int4), var_samp(b::int8), var_samp(b::numeric), + /* aggregate function = aggpartialfunc */ + any_value(b * 0), + bit_and(c_bit), bit_and(c_1or3int2), bit_and(c_1or3int4), bit_and(c_1or3int8), + bit_or(c_bit), bit_or(c_1or3int2), bit_or(c_1or3int4), bit_or(c_1or3int8), + bit_xor(c_bit), bit_xor(c_1or3int2), bit_xor(c_1or3int4), bit_xor(c_1or3int8), + bool_and(c_bool), + bool_or(c_bool), + count(b), count(*), + every(c_bool), + max(c_int4array), max(c_enum), max(c_1c::char(1)), max('2000-01-01'::date + b), max('0.0.0.0'::inet + b), max(b::float4), max(b::float8), max(b::int2), max(b::int4), max(b::int8), max(c_interval), max(c_money), max(b::numeric), max(b::oid), max(c_pg_lsn), max(c_tid), max(c_1c), max(c_time), max(c_timetz), max(c_timestamp), max(c_timestamptz), max(c_xid8), + min(c_int4array), min(c_enum), min(c_1c::char(1)), min('2000-01-01'::date + b), min('0.0.0.0'::inet + b), min(b::float4), min(b::float8), min(b::int2), min(b::int4), min(b::int8), min(c_interval), min(c_money), min(b::numeric), min(b::oid), min(c_pg_lsn), min(c_tid), min(c_1c), min(c_time), min(c_timetz), min(c_timestamp), min(c_timestamptz), min(c_xid8), + range_intersect_agg(c_int4range), range_intersect_agg(c_int4multirange), + regr_count((2 * b + 3)::float8, b::float8), + sum(b::float4), sum(b::float8), sum(b::int2), sum(b::int4), sum(c_interval), sum(c_money) + FROM pagg_tab WHERE c_serial between 1 and 30; + +-- Tests for backward compatibility +ALTER SERVER loopback OPTIONS (ADD check_partial_aggregate_support 'false'); +EXPLAIN (VERBOSE, COSTS OFF) +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; + +ALTER SERVER loopback OPTIONS (SET check_partial_aggregate_support 'true'); +EXPLAIN (VERBOSE, COSTS OFF) +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; + +ALTER SERVER loopback OPTIONS (DROP check_partial_aggregate_support); + +-- List of aggregate functions such that the aggregate function supports partial aggregate +-- and aggpartialfn field of the aggregate function is invalid +SELECT aggfnoid::text + FROM pg_aggregate a JOIN pg_type b ON a.aggtranstype = b.oid + WHERE (aggcombinefn::text <> '-') AND (aggpartialfn::text = '-') + ORDER BY 1; + +-- List of aggregate functions which have incorrect aggpartialfunc +SELECT a.aggfnoid + FROM pg_aggregate a JOIN pg_type b ON a.aggtranstype = b.oid + JOIN pg_aggregate c on a.aggpartialfn = c.aggfnoid + WHERE (a.aggcombinefn::text <> '-') + AND (a.aggpartialfn::text <> '-') + AND (a.aggtransfn <> c.aggtransfn + OR (c.aggcombinefn::text <> '-' + AND a.aggcombinefn <> c.aggcombinefn) + OR (a.agginitval IS NOT NULL + AND c.agginitval IS NOT NULL + AND a.agginitval <> c.agginitval) + OR (b.typname = 'internal' + AND c.aggserialfn::text <> '-' + AND a.aggserialfn <> c.aggserialfn) + OR (b.typname = 'internal' + AND c.aggdeserialfn::text <> '-' + AND a.aggdeserialfn <> c.aggdeserialfn) + OR (b.typname = 'internal' + AND a.aggserialfn <> c.aggfinalfn) + OR (b.typname <> 'internal' + AND c.aggfinalfn::text <> '-')) + ORDER BY 1; + +-- It's unsafe to push down partial aggregates which contain distinct clause +EXPLAIN (VERBOSE, COSTS OFF) +SELECT max(a), count(distinct b) FROM pagg_tab; +SELECT max(a), count(distinct b) FROM pagg_tab; + +-- It's unsafe to push down partial aggregates which contain order by clause +EXPLAIN (VERBOSE, COSTS OFF) SELECT array_agg(b order by b) FROM pagg_tab WHERE c_serial between 1 and 30; +SELECT array_agg(b order by b) FROM pagg_tab WHERE c_serial between 1 and 30; + +-- Create user-defined aggregates whose stype is internal +CREATE SCHEMA postgres_fdw_test; +CREATE AGGREGATE postgres_fdw_test.postgres_fdw_sum_p_int8(int8) ( + SFUNC = int8_avg_accum, + STYPE = internal, + COMBINEFUNC = int8_avg_combine, + FINALFUNC = int8_avg_serialize, + SERIALFUNC = int8_avg_serialize, + DESERIALFUNC = int8_avg_deserialize, + AGGPARTIALFUNC = postgres_fdw_test.postgres_fdw_sum_p_int8 +); + +CREATE AGGREGATE postgres_fdw_sum(int8) ( + SFUNC = int8_avg_accum, + STYPE = internal, + COMBINEFUNC = int8_avg_combine, + FINALFUNC = numeric_poly_sum, + SERIALFUNC = int8_avg_serialize, + DESERIALFUNC = int8_avg_deserialize, + AGGPARTIALFUNC = postgres_fdw_test.postgres_fdw_sum_p_int8 +); + +-- Create user-defined aggregates whose stype is not internal +CREATE AGGREGATE postgres_fdw_avg_p_int4(int4) ( + SFUNC = int4_avg_accum, + STYPE = _int8, + COMBINEFUNC = int4_avg_combine, + INITCOND = '{0,0}', + AGGPARTIALFUNC = postgres_fdw_avg_p_int4 +); + +CREATE AGGREGATE postgres_fdw_avg(int4) ( + SFUNC = int4_avg_accum, + STYPE = _int8, + COMBINEFUNC = int4_avg_combine, + FINALFUNC = int8_avg, + INITCOND = '{0,0}', + AGGPARTIALFUNC = postgres_fdw_avg_p_int4 +); + +CREATE AGGREGATE postgres_fdw_sum_noaggpartialfunc(int8) ( + SFUNC = int8_avg_accum, + STYPE = internal, + COMBINEFUNC = int8_avg_combine, + FINALFUNC = numeric_poly_sum, + SERIALFUNC = int8_avg_serialize, + DESERIALFUNC = int8_avg_deserialize +); + +-- It's safe to push down partial aggregate for user-defined aggregate +-- when it belongs to an extension that's listed in extensions set +-- and when its aggpartialfunc belongs to an extension that's listed +-- in extensions set. +ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_sum(int8); +ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_test.postgres_fdw_sum_p_int8(int8); + +ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_avg(int4); +ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_avg_p_int4(int4); + +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_sum(b::int8) FROM pagg_tab; +SELECT postgres_fdw_sum(b::int8) FROM pagg_tab; + +-- It's unsafe to push down partial aggregate for user-defined aggregate +-- when it has no aggpartialfunc. +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_sum_noaggpartialfunc(b::int8) FROM pagg_tab; +SELECT postgres_fdw_sum_noaggpartialfunc(b::int8) FROM pagg_tab; + +ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_test.postgres_fdw_sum_p_int8(int8); + +-- Reconnect for flushing the shippability cache +\c - +SET enable_partitionwise_aggregate TO true; + +-- It's unsafe to push down partial aggregate for user-defined aggregate +-- when itself is aggpartialfunc and when it belongs to an extension +-- that's listed in extensions set and when stype is internal. +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_test.postgres_fdw_sum_p_int8(b::int8) FROM pagg_tab; +SELECT postgres_fdw_test.postgres_fdw_sum_p_int8(b::int8) FROM pagg_tab; + +ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_sum(int8); +ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_test.postgres_fdw_sum_p_int8(int8); + +-- Reconnect for flushing the shippability cache +\c - +SET enable_partitionwise_aggregate TO true; + +-- It's unsafe to push down partial aggregate for user-defined aggregate +-- when it doesn't belong to an extension that's listed in extensions set +-- and when its aggpartialfunc belongs to an extension that's listed +-- in extensions set. +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_sum(b::int8) FROM pagg_tab; +SELECT postgres_fdw_sum(b::int8) FROM pagg_tab; + +-- It's safe to push down partial aggregate for user-defined aggregate +-- whose stype is not internal +-- when it belongs to an extension that's listed in extensions set +-- and when its aggpartialfunc belongs to an extension that's listed +-- in extensions set. +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_avg(b) FROM pagg_tab; +SELECT postgres_fdw_avg(b) FROM pagg_tab; + +-- It's safe to push down partial aggregate for user-defined aggregate +-- when itself is aggpartialfunc and when it belongs to an extension +-- that's listed in extensions set and when stype is not internal. +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_avg_p_int4(b) FROM pagg_tab; +SELECT postgres_fdw_avg_p_int4(b) FROM pagg_tab; + +ALTER EXTENSION postgres_fdw DROP TYPE mood; + +ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_test.postgres_fdw_sum_p_int8(int8); + +ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_avg_p_int4(int4); +ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_avg(int4); + +DROP AGGREGATE postgres_fdw_sum(int8); +DROP AGGREGATE postgres_fdw_test.postgres_fdw_sum_p_int8(int8); +DROP SCHEMA postgres_fdw_test; +DROP AGGREGATE postgres_fdw_avg(int4); +DROP AGGREGATE postgres_fdw_avg_p_int4(int4); +DROP AGGREGATE postgres_fdw_sum_noaggpartialfunc(int8); +ALTER SERVER loopback OPTIONS (DROP fdw_tuple_cost); + -- =================================================================== -- access rights and superuser -- =================================================================== diff --git a/doc/src/sgml/catalogs.sgml b/doc/src/sgml/catalogs.sgml index ed32ca0349..4f5ff32e96 100644 --- a/doc/src/sgml/catalogs.sgml +++ b/doc/src/sgml/catalogs.sgml @@ -644,6 +644,19 @@ value starts out null. </para></entry> </row> + + <row> + <entry role="catalog_table_entry"><para role="column_definition"> + <structfield>aggpartialfn</structfield> <type>regproc</type> + (references <link linkend="catalog-pg-proc"><structname>pg_proc</structname></link>.<structfield>oid</structfield>) + </para> + <para> + Partial aggregate function (zero if none). + See + <xref linkend="partial-aggregate-pushdown"/> for the + definition of partial aggregate function. + </para></entry> + </row> </tbody> </tgroup> </table> diff --git a/doc/src/sgml/postgres-fdw.sgml b/doc/src/sgml/postgres-fdw.sgml index 5062d712e7..676abc6fa5 100644 --- a/doc/src/sgml/postgres-fdw.sgml +++ b/doc/src/sgml/postgres-fdw.sgml @@ -451,6 +451,37 @@ OPTIONS (ADD password_required 'false'); </sect3> + <sect3 id="postgres-fdw-options-partial-aggregate-pushdown"> + <title>Partial Aggregate Pushdown Options</title> + + <para> + By default, <filename>postgres_fdw</filename> assumes that for each built-in aggregate + function, the partial aggregate function is defined on the remote server. See + <xref linkend="partial-aggregate-pushdown"/> for the definition of the partial aggregate + function. This may be overridden using the following option: + </para> + + <variablelist> + + <varlistentry> + <term><literal>check_partial_aggregate_support</literal> (<type>boolean</type>)</term> + <listitem> + <para> + When this option is true, <filename>postgres_fdw</filename> connects to the remote server + and gets its remote server version during query planning. + If the remote server version is older than the local server version, + <filename>postgres_fdw</filename> + assumes that for each aggregate function, the partial aggregate function is not defined + on the remote server unless the partial aggregate function and the aggregate + function match. + The default is <literal>false</literal>. + </para> + </listitem> + </varlistentry> + + </variablelist> + </sect3> + <sect3 id="postgres-fdw-options-asynchronous-execution"> <title>Asynchronous Execution Options</title> @@ -929,13 +960,15 @@ postgres=# SELECT postgres_fdw_disconnect_all(); <para> <filename>postgres_fdw</filename> attempts to optimize remote queries to reduce the amount of data transferred from foreign servers. This is done by - sending query <literal>WHERE</literal> clauses to the remote server for - execution, and by not retrieving table columns that are not needed for - the current query. To reduce the risk of misexecution of queries, - <literal>WHERE</literal> clauses are not sent to the remote server unless they use - only data types, operators, and functions that are built-in or belong to an - extension that's listed in the foreign server's <literal>extensions</literal> - option. Operators and functions in such clauses must + sending query <literal>WHERE</literal> clauses and aggregate expressions + to the remote server for execution, and by not retrieving table columns that + are not needed for the current query. + To reduce the risk of misexecution of queries, + <literal>WHERE</literal> clauses and aggregate expressions are not sent to + the remote server unless they use only data types, operators, and functions + that are built-in or belong to an extension that's listed in the foreign + server's <literal>extensions</literal> option. + Operators and functions in such clauses must be <literal>IMMUTABLE</literal> as well. For an <command>UPDATE</command> or <command>DELETE</command> query, <filename>postgres_fdw</filename> attempts to optimize the query execution by @@ -966,6 +999,68 @@ postgres=# SELECT postgres_fdw_disconnect_all(); </para> </sect2> + <sect2 id="partial-aggregate-pushdown"> + <title>Partial aggregate pushdown</title> + <para> + Partial aggregate pushdown is an optimization for a query that contains aggregate + expressions for a partitioned table across one or more remote servers. If the multiple + conditions are true for a remote server, an aggregate expression is sent to that + remote server and the aggregate is a partial aggregate on the local server. + The conditions under which this sending is active are as follows. + <itemizedlist spacing="compact"> + <listitem> + <para> + An aggregate function, called the partial aggregate function for partial aggregate + that corresponding to the aggregate expression, is defined on the local server and + the remote server. + </para> + </listitem> + <listitem> + <para> + The conditions in <xref linkend="postgres-fdw-remote-query-optimization"/> are true. + </para> + </listitem> + <listitem> + <para> + The query doesn't contain a <literal>HAVING</literal> clause. + </para> + </listitem> + <listitem> + <para> + The aggregate expressions in the query don't contain + a <literal>DISTINCT</literal> or <literal>ORDER BY</literal> clause. + </para> + </listitem> + </itemizedlist> + When this sending is active for aggregate expressions, any aggregate + function is replaced by the partial aggregate function in the remote query. + If the data type of the state value is not internal, the partial aggregate + function has no final function. + Otherwise, the partial aggregate function must have the same final function + as the aggregate function's serialization function. + The partial aggregate function and the aggregate function have + the same transition function and data type of the state value. + If the partial aggregate function has any of the following members: + initial state, combine function, serialization function, or deserialization + function, the members must match with the corresponding members in the + aggregate function. + For all built-in aggregate functions that support partial aggregation, the partial + aggregate function that corresponds to this function is defined as a built-in + aggregate function. + By default, <filename>postgres_fdw</filename> assumes that for each built-in + aggregate function, the partial aggregate function is defined on the local server + and the remote server. + If <literal>check_partial_aggregate_support</literal> is true and the remote server version is + older than the local server version, <filename>postgres_fdw</filename> does not assume + that the partial aggregate function is on the remote server unless + the partial aggregate function and the aggregate function match. + <filename>postgres_fdw</filename> assumes that a user-defined aggregate + function has the partial aggregate function on the remote server + only if both belong to an extension that is listed in the foreign server's + extensions option. + </para> + </sect2> + <sect2 id="postgres-fdw-remote-query-execution-environment"> <title>Remote Query Execution Environment</title> @@ -1031,14 +1126,27 @@ postgres=# SELECT postgres_fdw_disconnect_all(); back to 8.1. A limitation however is that <filename>postgres_fdw</filename> generally assumes that immutable built-in functions and operators are safe to send to the remote server for execution, if they appear in a - <literal>WHERE</literal> clause for a foreign table. Thus, a built-in - function that was added since the remote server's release might be sent - to it for execution, resulting in <quote>function does not exist</quote> or - a similar error. This type of failure can be worked around by + <literal>WHERE</literal> clause or aggregate expressions for a foreign table. + Thus, a built-in function that was added since the remote server's release + might be sent to it for execution, resulting in <quote>function does not + exist</quote> or a similar error. This type of failure can be worked around by rewriting the query, for example by embedding the foreign table reference in a sub-<literal>SELECT</literal> with <literal>OFFSET 0</literal> as an optimization fence, and placing the problematic function or operator outside the sub-<literal>SELECT</literal>. + A similar problem occurs when using the partial aggregate pushdown feature. + When using the partial aggregate pushdown feature, + the partial aggregate function of the aggregate function is sent to + the remote server if the conditions in + <xref linkend="partial-aggregate-pushdown"/> are met. + A limitation is that by default, <filename>postgres_fdw</filename> assumes that + for each built-in aggregate function, the partial aggregate function is defined + on the remote server. + This limitation might cause <quote>function does not exist</quote> + or a similar error if the remote server version is older than the local server version. + This problem can be worked around by setting + <literal>check_partial_aggregate_support</literal> to <literal>true</literal> + or similar query rewriting as above. </para> </sect2> diff --git a/doc/src/sgml/ref/create_aggregate.sgml b/doc/src/sgml/ref/create_aggregate.sgml index 222e0aa5c9..2875734dea 100644 --- a/doc/src/sgml/ref/create_aggregate.sgml +++ b/doc/src/sgml/ref/create_aggregate.sgml @@ -42,6 +42,7 @@ CREATE [ OR REPLACE ] AGGREGATE <replaceable class="parameter">name</replaceable [ , MINITCOND = <replaceable class="parameter">minitial_condition</replaceable> ] [ , SORTOP = <replaceable class="parameter">sort_operator</replaceable> ] [ , PARALLEL = { SAFE | RESTRICTED | UNSAFE } ] + [ , AGGPARTIALFUNC = <replaceable class="parameter">aggpartialfunc</replaceable> ] ) CREATE [ OR REPLACE ] AGGREGATE <replaceable class="parameter">name</replaceable> ( [ [ <replaceable class="parameter">argmode</replaceable> ] [ <replaceable class="parameter">argname</replaceable> ] <replaceable class="parameter">arg_data_type</replaceable> [ , ... ] ] @@ -80,6 +81,7 @@ CREATE [ OR REPLACE ] AGGREGATE <replaceable class="parameter">name</replaceable [ , MFINALFUNC_MODIFY = { READ_ONLY | SHAREABLE | READ_WRITE } ] [ , MINITCOND = <replaceable class="parameter">minitial_condition</replaceable> ] [ , SORTOP = <replaceable class="parameter">sort_operator</replaceable> ] + [ , AGGPARTIALFUNC = <replaceable class="parameter">aggpartialfunc</replaceable> ] ) </synopsis> </refsynopsisdiv> @@ -246,6 +248,18 @@ CREATE [ OR REPLACE ] AGGREGATE <replaceable class="parameter">name</replaceable marked <literal>PARALLEL SAFE</literal> to enable parallel aggregation. </para> + <para> + <filename>postgres_fdw</filename> can optimize a query containing aggregate + expressions for a partitioned table across one or more remote servers. + This optimization sends an aggregate expression to a remote server and the + aggregate is a partial aggregate on the local server. This optimization + requires an aggregate function called a <firstterm>partial aggregate + function</firstterm>. See + <xref linkend="partial-aggregate-pushdown"/> for the definition of + partial aggregate function. <literal>AGGPARTIALFUNC</literal> is a parameter + of the partial aggregate function. + </para> + <para> Aggregates that behave like <function>MIN</function> or <function>MAX</function> can sometimes be optimized by looking into an index instead of scanning every @@ -653,6 +667,34 @@ SELECT col FROM tab ORDER BY col USING sortop LIMIT 1; </para> </listitem> </varlistentry> + + <varlistentry> + <term><replaceable class="parameter">aggpartialfunc</replaceable></term> + <listitem> + <para> + The name of the partial aggregate function. + If the <replaceable class="parameter">name</replaceable> and the + <replaceable class="parameter">aggpartialfunc</replaceable> are the same, + then the aggregate function must be the partial aggregate function. + If the <replaceable class="parameter">aggpartialfunc</replaceable> + is specified, the <replaceable class="parameter">combinefunc</replaceable> + must be specified. + If the <replaceable class="parameter">state_data_type</replaceable> is not + <type>internal</type>, then the partial aggregate function must not have + the <replaceable class="parameter">ffunc</replaceable>. + If the <replaceable class="parameter">state_data_type</replaceable> is + <type>internal</type>, the partial aggregate function must have the same + final function as the <replaceable class="parameter">serialfunc</replaceable>. + The partial aggregate function and the aggregate function have + the same <replaceable class="parameter">sfunc</replaceable> and + <replaceable class="parameter">state_data_type</replaceable>. + If the partial aggregate function has any of the following members: + initial state, combine function, serialization function, or deserialization + function, the members must match with the corresponding members in the + aggregate function. + </para> + </listitem> + </varlistentry> </variablelist> <para> diff --git a/doc/src/sgml/xaggr.sgml b/doc/src/sgml/xaggr.sgml index bdad8d3dc2..607c859974 100644 --- a/doc/src/sgml/xaggr.sgml +++ b/doc/src/sgml/xaggr.sgml @@ -536,9 +536,6 @@ SELECT percentile_disc(0.5) WITHIN GROUP (ORDER BY income) FROM households; parallel aggregation by having different worker processes scan different portions of a table. Each worker produces a partial state value, and at the end those state values are combined to produce a final state value. - (In the future this mode might also be used for purposes such as combining - aggregations over local and remote tables; but that is not implemented - yet.) </para> <para> @@ -612,6 +609,18 @@ SELECT percentile_disc(0.5) WITHIN GROUP (ORDER BY income) FROM households; parallel-safety markings on its support functions are not consulted. </para> + <para> + <filename>postgres_fdw</filename> can optimize a query containing + aggregate expressions for a partitioned table across one or more + remote servers. This optimization sends an aggregate expression to + a remote server and the aggregate is a partial aggregate on the + local server. This optimization requires an aggregate function + called the <firstterm>partial aggregate function</firstterm>. + See + <xref linkend="partial-aggregate-pushdown"/> for the + definition of the partial aggregate function. + </para> + </sect2> <sect2 id="xaggr-support-functions"> diff --git a/src/backend/catalog/pg_aggregate.c b/src/backend/catalog/pg_aggregate.c index ebc4454743..aaf1e62762 100644 --- a/src/backend/catalog/pg_aggregate.c +++ b/src/backend/catalog/pg_aggregate.c @@ -63,6 +63,7 @@ AggregateCreate(const char *aggName, List *aggmtransfnName, List *aggminvtransfnName, List *aggmfinalfnName, + List *aggpartialfnName, bool finalfnExtraArgs, bool mfinalfnExtraArgs, char finalfnModify, @@ -91,6 +92,8 @@ AggregateCreate(const char *aggName, Oid mtransfn = InvalidOid; /* can be omitted */ Oid minvtransfn = InvalidOid; /* can be omitted */ Oid mfinalfn = InvalidOid; /* can be omitted */ + bool isaggpartialfn = false; + Oid aggpartialfn = InvalidOid; /* can be omitted */ Oid sortop = InvalidOid; /* can be omitted */ Oid *aggArgTypes = parameterTypes->values; bool mtransIsStrict = false; @@ -569,6 +572,94 @@ AggregateCreate(const char *aggName, format_type_be(finaltype)))); } + /* + * Validate the aggpartialfunc, if present. + */ + if (aggpartialfnName) + { + char *aggpartialName; + Oid aggpartialNamespace; + + if (!aggcombinefnName) + elog(ERROR, "aggcombinefnName must be supplied if aggpartialfnName is supplied"); + + /* Convert list of names to a name and namespace */ + aggpartialNamespace = QualifiedNameGetCreationNamespace(aggpartialfnName, + &aggpartialName); + + if ((aggNamespace == aggpartialNamespace) + && (strcmp(aggName, aggpartialName) == 0)) + { + if (((aggTransType != INTERNALOID) && (finalfn != InvalidOid)) + || ((aggTransType == INTERNALOID) && (finalfn != serialfn))) + elog(ERROR, "%s is not its own aggpartialfunc", aggName); + isaggpartialfn = true; + } + else + { + HeapTuple aggtup; + Form_pg_aggregate aggpartialform; + Datum textInitVal; + char *strInitVal; + bool initValueIsNull; + + aggpartialfn = LookupFuncName(aggpartialfnName, numArgs, aggArgTypes, false); + + /* Check aggregate creator has permission to call the function */ + aclresult = object_aclcheck(ProcedureRelationId, aggpartialfn, + GetUserId(), ACL_EXECUTE); + if (aclresult != ACLCHECK_OK) + aclcheck_error(aclresult, OBJECT_FUNCTION, get_func_name(aggpartialfn)); + + rettype = get_func_rettype(aggpartialfn); + + aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(aggpartialfn)); + if (!HeapTupleIsValid(aggtup)) + elog(ERROR, "cache lookup failed for aggpartialfunc %u", aggpartialfn); + + aggpartialform = (Form_pg_aggregate) GETSTRUCT(aggtup); + + if ((aggpartialform->aggcombinefn != InvalidOid) && + (aggpartialform->aggcombinefn != combinefn)) + elog(ERROR, "%s in aggregate and %s in its aggpartialfunc must match", + "combinefunc", "combinefunc"); + + if (aggpartialform->aggtransfn != transfn) + elog(ERROR, "%s in aggregate and its aggpartialfunc must match", "sfunc"); + + textInitVal = SysCacheGetAttr(AGGFNOID, aggtup, + Anum_pg_aggregate_agginitval, &initValueIsNull); + + if (!initValueIsNull && (agginitval != NULL)) + { + strInitVal = TextDatumGetCString(textInitVal); + if (strcmp(strInitVal, agginitval) != 0) + elog(ERROR, "%s in aggregate and %s in its aggpartialfunc must match", + "initcond", "initcond"); + } + + if (aggTransType == INTERNALOID) + { + if (aggpartialform->aggcombinefn != InvalidOid) + { + if (aggpartialform->aggserialfn != serialfn) + elog(ERROR, "%s in aggregate and %s in its aggpartialfunc must match", + "serialfunc", "serialfunc"); + + if (aggpartialform->aggdeserialfn != deserialfn) + elog(ERROR, "%s in aggregate and %s in its aggpartialfunc must match", + "deserialfunc", "deserialfunc"); + } + if (aggpartialform->aggfinalfn != serialfn) + elog(ERROR, "finalfunc of aggpartialfunc must match serialfunc of aggregate when stype is internal"); + } + else if (aggpartialform->aggfinalfn != InvalidOid) + elog(ERROR, "finalfunc of aggpartialfunc must not be supplied when stype isn't internal"); + + ReleaseSysCache(aggtup); + } + } + /* handle sortop, if supplied */ if (aggsortopName) { @@ -684,6 +775,9 @@ AggregateCreate(const char *aggName, values[Anum_pg_aggregate_aggminitval - 1] = CStringGetTextDatum(aggminitval); else nulls[Anum_pg_aggregate_aggminitval - 1] = true; + if (isaggpartialfn) + aggpartialfn = procOid; + values[Anum_pg_aggregate_aggpartialfn - 1] = ObjectIdGetDatum(aggpartialfn); if (replace) oldtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(procOid)); @@ -805,6 +899,13 @@ AggregateCreate(const char *aggName, add_exact_object_address(&referenced, addrs); } + /* Depends on aggpartialfunc, if any */ + if (OidIsValid(aggpartialfn) && (aggpartialfn != procOid)) + { + ObjectAddressSet(referenced, ProcedureRelationId, aggpartialfn); + add_exact_object_address(&referenced, addrs); + } + record_object_address_dependencies(&myself, addrs, DEPENDENCY_NORMAL); free_object_addresses(addrs); return myself; diff --git a/src/backend/commands/aggregatecmds.c b/src/backend/commands/aggregatecmds.c index fda9d1aa77..8c8f1789e9 100644 --- a/src/backend/commands/aggregatecmds.c +++ b/src/backend/commands/aggregatecmds.c @@ -70,6 +70,7 @@ DefineAggregate(ParseState *pstate, List *combinefuncName = NIL; List *serialfuncName = NIL; List *deserialfuncName = NIL; + List *aggpartialfuncName = NIL; List *mtransfuncName = NIL; List *minvtransfuncName = NIL; List *mfinalfuncName = NIL; @@ -143,6 +144,8 @@ DefineAggregate(ParseState *pstate, serialfuncName = defGetQualifiedName(defel); else if (strcmp(defel->defname, "deserialfunc") == 0) deserialfuncName = defGetQualifiedName(defel); + else if (strcmp(defel->defname, "aggpartialfunc") == 0) + aggpartialfuncName = defGetQualifiedName(defel); else if (strcmp(defel->defname, "msfunc") == 0) mtransfuncName = defGetQualifiedName(defel); else if (strcmp(defel->defname, "minvfunc") == 0) @@ -461,6 +464,7 @@ DefineAggregate(ParseState *pstate, mtransfuncName, /* fwd trans function name */ minvtransfuncName, /* inv trans function name */ mfinalfuncName, /* final function name */ + aggpartialfuncName, finalfuncExtraArgs, mfinalfuncExtraArgs, finalfuncModify, diff --git a/src/bin/pg_dump/pg_dump.c b/src/bin/pg_dump/pg_dump.c index 5dab1ba9ea..b958145ec2 100644 --- a/src/bin/pg_dump/pg_dump.c +++ b/src/bin/pg_dump/pg_dump.c @@ -13679,6 +13679,7 @@ dumpAgg(Archive *fout, const AggInfo *agginfo) const char *aggmtransfn; const char *aggminvtransfn; const char *aggmfinalfn; + const char *aggpartialfn; bool aggfinalextra; bool aggmfinalextra; char aggfinalmodify; @@ -13761,11 +13762,11 @@ dumpAgg(Archive *fout, const AggInfo *agginfo) if (fout->remoteVersion >= 110000) appendPQExpBufferStr(query, "aggfinalmodify,\n" - "aggmfinalmodify\n"); + "aggmfinalmodify,\n"); else appendPQExpBufferStr(query, "'0' AS aggfinalmodify,\n" - "'0' AS aggmfinalmodify\n"); + "'0' AS aggmfinalmodify,\n"); appendPQExpBufferStr(query, "FROM pg_catalog.pg_aggregate a, pg_catalog.pg_proc p " @@ -13791,6 +13792,7 @@ dumpAgg(Archive *fout, const AggInfo *agginfo) aggcombinefn = PQgetvalue(res, 0, PQfnumber(res, "aggcombinefn")); aggserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggserialfn")); aggdeserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggdeserialfn")); + aggpartialfn = PQgetvalue(res, 0, PQfnumber(res, "aggpartialfn")); aggmtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggmtransfn")); aggminvtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggminvtransfn")); aggmfinalfn = PQgetvalue(res, 0, PQfnumber(res, "aggmfinalfn")); @@ -13880,6 +13882,9 @@ dumpAgg(Archive *fout, const AggInfo *agginfo) if (strcmp(aggdeserialfn, "-") != 0) appendPQExpBuffer(details, ",\n DESERIALFUNC = %s", aggdeserialfn); + if (strcmp(aggpartialfn, "-") != 0) + appendPQExpBuffer(details, ",\n AGGPARTIALFUNC = %s", aggpartialfn); + if (strcmp(aggmtransfn, "-") != 0) { appendPQExpBuffer(details, ",\n MSFUNC = %s,\n MINVFUNC = %s,\n MSTYPE = %s", diff --git a/src/include/catalog/pg_aggregate.dat b/src/include/catalog/pg_aggregate.dat index 1bc1d97d74..dd97b4dc34 100644 --- a/src/include/catalog/pg_aggregate.dat +++ b/src/include/catalog/pg_aggregate.dat @@ -13,22 +13,44 @@ [ # avg +{ aggfnoid => 'avg_p_int8', aggtransfn => 'int8_avg_accum', + aggfinalfn => 'int8_avg_serialize', aggcombinefn => 'int8_avg_combine', + aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize', + aggtranstype => 'internal', aggtransspace => '48', + aggpartialfn => 'avg_p_int8' }, { aggfnoid => 'avg(int8)', aggtransfn => 'int8_avg_accum', aggfinalfn => 'numeric_poly_avg', aggcombinefn => 'int8_avg_combine', aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize', aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv', aggmfinalfn => 'numeric_poly_avg', aggtranstype => 'internal', - aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' }, + aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'avg_p_int8' }, +{ aggfnoid => 'avg_p_int4', aggtransfn => 'int4_avg_accum', + aggcombinefn => 'int4_avg_combine', aggtranstype => '_int8', + agginitval => '{0,0}', + aggpartialfn => 'avg_p_int4' }, { aggfnoid => 'avg(int4)', aggtransfn => 'int4_avg_accum', aggfinalfn => 'int8_avg', aggcombinefn => 'int4_avg_combine', aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv', aggmfinalfn => 'int8_avg', aggtranstype => '_int8', aggmtranstype => '_int8', - agginitval => '{0,0}', aggminitval => '{0,0}' }, + agginitval => '{0,0}', aggminitval => '{0,0}', + aggpartialfn => 'avg_p_int4' }, +{ aggfnoid => 'avg_p_int2', aggtransfn => 'int2_avg_accum', + aggcombinefn => 'int4_avg_combine', aggtranstype => '_int8', + agginitval => '{0,0}', + aggpartialfn => 'avg_p_int2' }, { aggfnoid => 'avg(int2)', aggtransfn => 'int2_avg_accum', aggfinalfn => 'int8_avg', aggcombinefn => 'int4_avg_combine', aggmtransfn => 'int2_avg_accum', aggminvtransfn => 'int2_avg_accum_inv', aggmfinalfn => 'int8_avg', aggtranstype => '_int8', aggmtranstype => '_int8', - agginitval => '{0,0}', aggminitval => '{0,0}' }, + agginitval => '{0,0}', aggminitval => '{0,0}', + aggpartialfn => 'avg_p_int2' }, +{ aggfnoid => 'avg_p_numeric', aggtransfn => 'numeric_avg_accum', + aggfinalfn => 'numeric_avg_serialize', aggcombinefn => 'numeric_avg_combine', + aggserialfn => 'numeric_avg_serialize', + aggdeserialfn => 'numeric_avg_deserialize', aggtranstype => 'internal', + aggtransspace => '128', + aggpartialfn => 'avg_p_numeric' }, { aggfnoid => 'avg(numeric)', aggtransfn => 'numeric_avg_accum', aggfinalfn => 'numeric_avg', aggcombinefn => 'numeric_avg_combine', aggserialfn => 'numeric_avg_serialize', @@ -36,46 +58,80 @@ aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv', aggmfinalfn => 'numeric_avg', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'avg_p_numeric' }, +{ aggfnoid => 'avg_p_float4', aggtransfn => 'float4_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'avg_p_float4' }, { aggfnoid => 'avg(float4)', aggtransfn => 'float4_accum', aggfinalfn => 'float8_avg', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'avg_p_float4' }, +{ aggfnoid => 'avg_p_float8', aggtransfn => 'float8_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'avg_p_float8' }, { aggfnoid => 'avg(float8)', aggtransfn => 'float8_accum', aggfinalfn => 'float8_avg', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'avg_p_float8' }, +{ aggfnoid => 'avg_p_interval', aggtransfn => 'interval_accum', + aggcombinefn => 'interval_combine', aggtranstype => '_interval', + agginitval => '{0 second,0 second}', + aggpartialfn => 'avg_p_interval' }, { aggfnoid => 'avg(interval)', aggtransfn => 'interval_accum', aggfinalfn => 'interval_avg', aggcombinefn => 'interval_combine', aggmtransfn => 'interval_accum', aggminvtransfn => 'interval_accum_inv', aggmfinalfn => 'interval_avg', aggtranstype => '_interval', aggmtranstype => '_interval', agginitval => '{0 second,0 second}', - aggminitval => '{0 second,0 second}' }, + aggminitval => '{0 second,0 second}', + aggpartialfn => 'avg_p_interval' }, # sum +{ aggfnoid => 'sum_p_int8', aggtransfn => 'int8_avg_accum', + aggfinalfn => 'int8_avg_serialize', aggcombinefn => 'int8_avg_combine', + aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize', + aggtranstype => 'internal', aggtransspace => '48', + aggpartialfn => 'sum_p_int8' }, { aggfnoid => 'sum(int8)', aggtransfn => 'int8_avg_accum', aggfinalfn => 'numeric_poly_sum', aggcombinefn => 'int8_avg_combine', aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize', aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv', aggmfinalfn => 'numeric_poly_sum', aggtranstype => 'internal', - aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' }, + aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'sum_p_int8' }, { aggfnoid => 'sum(int4)', aggtransfn => 'int4_sum', aggcombinefn => 'int8pl', aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv', aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8', - aggmtranstype => '_int8', aggminitval => '{0,0}' }, + aggmtranstype => '_int8', aggminitval => '{0,0}', + aggpartialfn => 'sum(int4)' }, { aggfnoid => 'sum(int2)', aggtransfn => 'int2_sum', aggcombinefn => 'int8pl', aggmtransfn => 'int2_avg_accum', aggminvtransfn => 'int2_avg_accum_inv', aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8', - aggmtranstype => '_int8', aggminitval => '{0,0}' }, + aggmtranstype => '_int8', aggminitval => '{0,0}', + aggpartialfn => 'sum(int2)' }, { aggfnoid => 'sum(float4)', aggtransfn => 'float4pl', - aggcombinefn => 'float4pl', aggtranstype => 'float4' }, + aggcombinefn => 'float4pl', aggtranstype => 'float4', + aggpartialfn => 'sum(float4)' }, { aggfnoid => 'sum(float8)', aggtransfn => 'float8pl', - aggcombinefn => 'float8pl', aggtranstype => 'float8' }, + aggcombinefn => 'float8pl', aggtranstype => 'float8', + aggpartialfn => 'sum(float8)' }, { aggfnoid => 'sum(money)', aggtransfn => 'cash_pl', aggcombinefn => 'cash_pl', aggmtransfn => 'cash_pl', aggminvtransfn => 'cash_mi', - aggtranstype => 'money', aggmtranstype => 'money' }, + aggtranstype => 'money', aggmtranstype => 'money', + aggpartialfn => 'sum(money)' }, { aggfnoid => 'sum(interval)', aggtransfn => 'interval_pl', aggcombinefn => 'interval_pl', aggmtransfn => 'interval_pl', aggminvtransfn => 'interval_mi', aggtranstype => 'interval', - aggmtranstype => 'interval' }, + aggmtranstype => 'interval', + aggpartialfn => 'sum(interval)' }, +{ aggfnoid => 'sum_p_numeric', aggtransfn => 'numeric_avg_accum', + aggfinalfn => 'numeric_avg_serialize', aggcombinefn => 'numeric_avg_combine', + aggserialfn => 'numeric_avg_serialize', + aggdeserialfn => 'numeric_avg_deserialize', aggtranstype => 'internal', + aggtransspace => '128', + aggpartialfn => 'sum_p_numeric' }, { aggfnoid => 'sum(numeric)', aggtransfn => 'numeric_avg_accum', aggfinalfn => 'numeric_sum', aggcombinefn => 'numeric_avg_combine', aggserialfn => 'numeric_avg_serialize', @@ -83,269 +139,436 @@ aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv', aggmfinalfn => 'numeric_sum', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'sum_p_numeric' }, # max { aggfnoid => 'max(int8)', aggtransfn => 'int8larger', aggcombinefn => 'int8larger', aggsortop => '>(int8,int8)', - aggtranstype => 'int8' }, + aggtranstype => 'int8', + aggpartialfn => 'max(int8)' }, { aggfnoid => 'max(int4)', aggtransfn => 'int4larger', aggcombinefn => 'int4larger', aggsortop => '>(int4,int4)', - aggtranstype => 'int4' }, + aggtranstype => 'int4', + aggpartialfn => 'max(int4)' }, { aggfnoid => 'max(int2)', aggtransfn => 'int2larger', aggcombinefn => 'int2larger', aggsortop => '>(int2,int2)', - aggtranstype => 'int2' }, + aggtranstype => 'int2', + aggpartialfn => 'max(int2)' }, { aggfnoid => 'max(oid)', aggtransfn => 'oidlarger', aggcombinefn => 'oidlarger', aggsortop => '>(oid,oid)', - aggtranstype => 'oid' }, + aggtranstype => 'oid', + aggpartialfn => 'max(oid)' }, { aggfnoid => 'max(float4)', aggtransfn => 'float4larger', aggcombinefn => 'float4larger', aggsortop => '>(float4,float4)', - aggtranstype => 'float4' }, + aggtranstype => 'float4', + aggpartialfn => 'max(float4)' }, { aggfnoid => 'max(float8)', aggtransfn => 'float8larger', aggcombinefn => 'float8larger', aggsortop => '>(float8,float8)', - aggtranstype => 'float8' }, + aggtranstype => 'float8', + aggpartialfn => 'max(float8)' }, { aggfnoid => 'max(date)', aggtransfn => 'date_larger', aggcombinefn => 'date_larger', aggsortop => '>(date,date)', - aggtranstype => 'date' }, + aggtranstype => 'date', + aggpartialfn => 'max(date)' }, { aggfnoid => 'max(time)', aggtransfn => 'time_larger', aggcombinefn => 'time_larger', aggsortop => '>(time,time)', - aggtranstype => 'time' }, + aggtranstype => 'time', + aggpartialfn => 'max(time)' }, { aggfnoid => 'max(timetz)', aggtransfn => 'timetz_larger', aggcombinefn => 'timetz_larger', aggsortop => '>(timetz,timetz)', - aggtranstype => 'timetz' }, + aggtranstype => 'timetz', + aggpartialfn => 'max(timetz)' }, { aggfnoid => 'max(money)', aggtransfn => 'cashlarger', aggcombinefn => 'cashlarger', aggsortop => '>(money,money)', - aggtranstype => 'money' }, + aggtranstype => 'money', + aggpartialfn => 'max(money)' }, { aggfnoid => 'max(timestamp)', aggtransfn => 'timestamp_larger', aggcombinefn => 'timestamp_larger', aggsortop => '>(timestamp,timestamp)', - aggtranstype => 'timestamp' }, + aggtranstype => 'timestamp', + aggpartialfn => 'max(timestamp)' }, { aggfnoid => 'max(timestamptz)', aggtransfn => 'timestamptz_larger', aggcombinefn => 'timestamptz_larger', - aggsortop => '>(timestamptz,timestamptz)', aggtranstype => 'timestamptz' }, + aggsortop => '>(timestamptz,timestamptz)', aggtranstype => 'timestamptz', + aggpartialfn => 'max(timestamptz)' }, { aggfnoid => 'max(interval)', aggtransfn => 'interval_larger', aggcombinefn => 'interval_larger', aggsortop => '>(interval,interval)', - aggtranstype => 'interval' }, + aggtranstype => 'interval', + aggpartialfn => 'max(interval)' }, { aggfnoid => 'max(text)', aggtransfn => 'text_larger', aggcombinefn => 'text_larger', aggsortop => '>(text,text)', - aggtranstype => 'text' }, + aggtranstype => 'text', + aggpartialfn => 'max(text)' }, { aggfnoid => 'max(numeric)', aggtransfn => 'numeric_larger', aggcombinefn => 'numeric_larger', aggsortop => '>(numeric,numeric)', - aggtranstype => 'numeric' }, + aggtranstype => 'numeric', + aggpartialfn => 'max(numeric)' }, { aggfnoid => 'max(anyarray)', aggtransfn => 'array_larger', aggcombinefn => 'array_larger', aggsortop => '>(anyarray,anyarray)', - aggtranstype => 'anyarray' }, + aggtranstype => 'anyarray', + aggpartialfn => 'max(anyarray)' }, { aggfnoid => 'max(bpchar)', aggtransfn => 'bpchar_larger', aggcombinefn => 'bpchar_larger', aggsortop => '>(bpchar,bpchar)', - aggtranstype => 'bpchar' }, + aggtranstype => 'bpchar', + aggpartialfn => 'max(bpchar)' }, { aggfnoid => 'max(tid)', aggtransfn => 'tidlarger', aggcombinefn => 'tidlarger', aggsortop => '>(tid,tid)', - aggtranstype => 'tid' }, + aggtranstype => 'tid', + aggpartialfn => 'max(tid)' }, { aggfnoid => 'max(anyenum)', aggtransfn => 'enum_larger', aggcombinefn => 'enum_larger', aggsortop => '>(anyenum,anyenum)', - aggtranstype => 'anyenum' }, + aggtranstype => 'anyenum', + aggpartialfn => 'max(anyenum)' }, { aggfnoid => 'max(inet)', aggtransfn => 'network_larger', aggcombinefn => 'network_larger', aggsortop => '>(inet,inet)', - aggtranstype => 'inet' }, + aggtranstype => 'inet', + aggpartialfn => 'max(inet)' }, { aggfnoid => 'max(pg_lsn)', aggtransfn => 'pg_lsn_larger', aggcombinefn => 'pg_lsn_larger', aggsortop => '>(pg_lsn,pg_lsn)', - aggtranstype => 'pg_lsn' }, + aggtranstype => 'pg_lsn', + aggpartialfn => 'max(pg_lsn)' }, { aggfnoid => 'max(xid8)', aggtransfn => 'xid8_larger', aggcombinefn => 'xid8_larger', aggsortop => '>(xid8,xid8)', - aggtranstype => 'xid8' }, + aggtranstype => 'xid8', + aggpartialfn => 'max(xid8)' }, # min { aggfnoid => 'min(int8)', aggtransfn => 'int8smaller', aggcombinefn => 'int8smaller', aggsortop => '<(int8,int8)', - aggtranstype => 'int8' }, + aggtranstype => 'int8', + aggpartialfn => 'min(int8)' }, { aggfnoid => 'min(int4)', aggtransfn => 'int4smaller', aggcombinefn => 'int4smaller', aggsortop => '<(int4,int4)', - aggtranstype => 'int4' }, + aggtranstype => 'int4', + aggpartialfn => 'min(int4)' }, { aggfnoid => 'min(int2)', aggtransfn => 'int2smaller', aggcombinefn => 'int2smaller', aggsortop => '<(int2,int2)', - aggtranstype => 'int2' }, + aggtranstype => 'int2', + aggpartialfn => 'min(int2)' }, { aggfnoid => 'min(oid)', aggtransfn => 'oidsmaller', aggcombinefn => 'oidsmaller', aggsortop => '<(oid,oid)', - aggtranstype => 'oid' }, + aggtranstype => 'oid', + aggpartialfn => 'min(oid)' }, { aggfnoid => 'min(float4)', aggtransfn => 'float4smaller', aggcombinefn => 'float4smaller', aggsortop => '<(float4,float4)', - aggtranstype => 'float4' }, + aggtranstype => 'float4', + aggpartialfn => 'min(float4)' }, { aggfnoid => 'min(float8)', aggtransfn => 'float8smaller', aggcombinefn => 'float8smaller', aggsortop => '<(float8,float8)', - aggtranstype => 'float8' }, + aggtranstype => 'float8', + aggpartialfn => 'min(float8)' }, { aggfnoid => 'min(date)', aggtransfn => 'date_smaller', aggcombinefn => 'date_smaller', aggsortop => '<(date,date)', - aggtranstype => 'date' }, + aggtranstype => 'date', + aggpartialfn => 'min(date)' }, { aggfnoid => 'min(time)', aggtransfn => 'time_smaller', aggcombinefn => 'time_smaller', aggsortop => '<(time,time)', - aggtranstype => 'time' }, + aggtranstype => 'time', + aggpartialfn => 'min(time)' }, { aggfnoid => 'min(timetz)', aggtransfn => 'timetz_smaller', aggcombinefn => 'timetz_smaller', aggsortop => '<(timetz,timetz)', - aggtranstype => 'timetz' }, + aggtranstype => 'timetz', + aggpartialfn => 'min(timetz)' }, { aggfnoid => 'min(money)', aggtransfn => 'cashsmaller', aggcombinefn => 'cashsmaller', aggsortop => '<(money,money)', - aggtranstype => 'money' }, + aggtranstype => 'money', + aggpartialfn => 'min(money)' }, { aggfnoid => 'min(timestamp)', aggtransfn => 'timestamp_smaller', aggcombinefn => 'timestamp_smaller', aggsortop => '<(timestamp,timestamp)', - aggtranstype => 'timestamp' }, + aggtranstype => 'timestamp', + aggpartialfn => 'min(timestamp)' }, { aggfnoid => 'min(timestamptz)', aggtransfn => 'timestamptz_smaller', aggcombinefn => 'timestamptz_smaller', - aggsortop => '<(timestamptz,timestamptz)', aggtranstype => 'timestamptz' }, + aggsortop => '<(timestamptz,timestamptz)', aggtranstype => 'timestamptz', + aggpartialfn => 'min(timestamptz)' }, { aggfnoid => 'min(interval)', aggtransfn => 'interval_smaller', aggcombinefn => 'interval_smaller', aggsortop => '<(interval,interval)', - aggtranstype => 'interval' }, + aggtranstype => 'interval', + aggpartialfn => 'min(interval)' }, { aggfnoid => 'min(text)', aggtransfn => 'text_smaller', aggcombinefn => 'text_smaller', aggsortop => '<(text,text)', - aggtranstype => 'text' }, + aggtranstype => 'text', + aggpartialfn => 'min(text)' }, { aggfnoid => 'min(numeric)', aggtransfn => 'numeric_smaller', aggcombinefn => 'numeric_smaller', aggsortop => '<(numeric,numeric)', - aggtranstype => 'numeric' }, + aggtranstype => 'numeric', + aggpartialfn => 'min(numeric)' }, { aggfnoid => 'min(anyarray)', aggtransfn => 'array_smaller', aggcombinefn => 'array_smaller', aggsortop => '<(anyarray,anyarray)', - aggtranstype => 'anyarray' }, + aggtranstype => 'anyarray', + aggpartialfn => 'min(anyarray)' }, { aggfnoid => 'min(bpchar)', aggtransfn => 'bpchar_smaller', aggcombinefn => 'bpchar_smaller', aggsortop => '<(bpchar,bpchar)', - aggtranstype => 'bpchar' }, + aggtranstype => 'bpchar', + aggpartialfn => 'min(bpchar)' }, { aggfnoid => 'min(tid)', aggtransfn => 'tidsmaller', aggcombinefn => 'tidsmaller', aggsortop => '<(tid,tid)', - aggtranstype => 'tid' }, + aggtranstype => 'tid', + aggpartialfn => 'min(tid)' }, { aggfnoid => 'min(anyenum)', aggtransfn => 'enum_smaller', aggcombinefn => 'enum_smaller', aggsortop => '<(anyenum,anyenum)', - aggtranstype => 'anyenum' }, + aggtranstype => 'anyenum', + aggpartialfn => 'min(anyenum)' }, { aggfnoid => 'min(inet)', aggtransfn => 'network_smaller', aggcombinefn => 'network_smaller', aggsortop => '<(inet,inet)', - aggtranstype => 'inet' }, + aggtranstype => 'inet', + aggpartialfn => 'min(inet)' }, { aggfnoid => 'min(pg_lsn)', aggtransfn => 'pg_lsn_smaller', aggcombinefn => 'pg_lsn_smaller', aggsortop => '<(pg_lsn,pg_lsn)', - aggtranstype => 'pg_lsn' }, + aggtranstype => 'pg_lsn', + aggpartialfn => 'min(pg_lsn)' }, { aggfnoid => 'min(xid8)', aggtransfn => 'xid8_smaller', aggcombinefn => 'xid8_smaller', aggsortop => '<(xid8,xid8)', - aggtranstype => 'xid8' }, + aggtranstype => 'xid8', + aggpartialfn => 'min(xid8)' }, # count { aggfnoid => 'count(any)', aggtransfn => 'int8inc_any', aggcombinefn => 'int8pl', aggmtransfn => 'int8inc_any', aggminvtransfn => 'int8dec_any', aggtranstype => 'int8', - aggmtranstype => 'int8', agginitval => '0', aggminitval => '0' }, + aggmtranstype => 'int8', agginitval => '0', aggminitval => '0', + aggpartialfn => 'count(any)' }, { aggfnoid => 'count()', aggtransfn => 'int8inc', aggcombinefn => 'int8pl', aggmtransfn => 'int8inc', aggminvtransfn => 'int8dec', aggtranstype => 'int8', - aggmtranstype => 'int8', agginitval => '0', aggminitval => '0' }, + aggmtranstype => 'int8', agginitval => '0', aggminitval => '0', + aggpartialfn => 'count()' }, # var_pop +{ aggfnoid => 'var_pop_p_int8', aggtransfn => 'int8_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'var_pop_p_int8' }, { aggfnoid => 'var_pop(int8)', aggtransfn => 'int8_accum', aggfinalfn => 'numeric_var_pop', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'int8_accum', aggminvtransfn => 'int8_accum_inv', aggmfinalfn => 'numeric_var_pop', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'var_pop_p_int8' }, +{ aggfnoid => 'var_pop_p_int4', aggtransfn => 'int4_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'var_pop_p_int4' }, { aggfnoid => 'var_pop(int4)', aggtransfn => 'int4_accum', aggfinalfn => 'numeric_poly_var_pop', aggcombinefn => 'numeric_poly_combine', aggserialfn => 'numeric_poly_serialize', aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int4_accum', aggminvtransfn => 'int4_accum_inv', aggmfinalfn => 'numeric_poly_var_pop', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'var_pop_p_int4' }, +{ aggfnoid => 'var_pop_p_int2', aggtransfn => 'int2_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'var_pop_p_int2' }, { aggfnoid => 'var_pop(int2)', aggtransfn => 'int2_accum', aggfinalfn => 'numeric_poly_var_pop', aggcombinefn => 'numeric_poly_combine', aggserialfn => 'numeric_poly_serialize', aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int2_accum', aggminvtransfn => 'int2_accum_inv', aggmfinalfn => 'numeric_poly_var_pop', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'var_pop_p_int2' }, +{ aggfnoid => 'var_pop_p_float4', aggtransfn => 'float4_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'var_pop_p_float4' }, { aggfnoid => 'var_pop(float4)', aggtransfn => 'float4_accum', aggfinalfn => 'float8_var_pop', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'var_pop_p_float4' }, +{ aggfnoid => 'var_pop_p_float8', aggtransfn => 'float8_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'var_pop_p_float8' }, { aggfnoid => 'var_pop(float8)', aggtransfn => 'float8_accum', aggfinalfn => 'float8_var_pop', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'var_pop_p_float8' }, +{ aggfnoid => 'var_pop_p_numeric', aggtransfn => 'numeric_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'var_pop_p_numeric' }, { aggfnoid => 'var_pop(numeric)', aggtransfn => 'numeric_accum', aggfinalfn => 'numeric_var_pop', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'numeric_accum', aggminvtransfn => 'numeric_accum_inv', aggmfinalfn => 'numeric_var_pop', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'var_pop_p_numeric' }, # var_samp +{ aggfnoid => 'var_samp_p_int8', aggtransfn => 'int8_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'var_samp_p_int8' }, { aggfnoid => 'var_samp(int8)', aggtransfn => 'int8_accum', aggfinalfn => 'numeric_var_samp', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'int8_accum', aggminvtransfn => 'int8_accum_inv', aggmfinalfn => 'numeric_var_samp', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'var_samp_p_int8' }, +{ aggfnoid => 'var_samp_p_int4', aggtransfn => 'int4_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'var_samp_p_int4' }, { aggfnoid => 'var_samp(int4)', aggtransfn => 'int4_accum', aggfinalfn => 'numeric_poly_var_samp', aggcombinefn => 'numeric_poly_combine', aggserialfn => 'numeric_poly_serialize', aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int4_accum', aggminvtransfn => 'int4_accum_inv', aggmfinalfn => 'numeric_poly_var_samp', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'var_samp_p_int4' }, +{ aggfnoid => 'var_samp_p_int2', aggtransfn => 'int2_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'var_samp_p_int2' }, { aggfnoid => 'var_samp(int2)', aggtransfn => 'int2_accum', aggfinalfn => 'numeric_poly_var_samp', aggcombinefn => 'numeric_poly_combine', aggserialfn => 'numeric_poly_serialize', aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int2_accum', aggminvtransfn => 'int2_accum_inv', aggmfinalfn => 'numeric_poly_var_samp', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'var_samp_p_int2' }, +{ aggfnoid => 'var_samp_p_float4', aggtransfn => 'float4_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'var_samp_p_float4' }, { aggfnoid => 'var_samp(float4)', aggtransfn => 'float4_accum', aggfinalfn => 'float8_var_samp', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'var_samp_p_float4' }, +{ aggfnoid => 'var_samp_p_float8', aggtransfn => 'float8_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'var_samp_p_float8' }, { aggfnoid => 'var_samp(float8)', aggtransfn => 'float8_accum', aggfinalfn => 'float8_var_samp', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'var_samp_p_float8' }, +{ aggfnoid => 'var_samp_p_numeric', aggtransfn => 'numeric_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'var_samp_p_numeric' }, { aggfnoid => 'var_samp(numeric)', aggtransfn => 'numeric_accum', aggfinalfn => 'numeric_var_samp', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'numeric_accum', aggminvtransfn => 'numeric_accum_inv', aggmfinalfn => 'numeric_var_samp', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'var_samp_p_numeric' }, # variance: historical Postgres syntax for var_samp +{ aggfnoid => 'variance_p_int8', aggtransfn => 'int8_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'variance_p_int8' }, { aggfnoid => 'variance(int8)', aggtransfn => 'int8_accum', aggfinalfn => 'numeric_var_samp', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'int8_accum', aggminvtransfn => 'int8_accum_inv', aggmfinalfn => 'numeric_var_samp', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'variance_p_int8' }, +{ aggfnoid => 'variance_p_int4', aggtransfn => 'int4_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'variance_p_int4' }, { aggfnoid => 'variance(int4)', aggtransfn => 'int4_accum', aggfinalfn => 'numeric_poly_var_samp', aggcombinefn => 'numeric_poly_combine', aggserialfn => 'numeric_poly_serialize', aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int4_accum', aggminvtransfn => 'int4_accum_inv', aggmfinalfn => 'numeric_poly_var_samp', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'variance_p_int4' }, +{ aggfnoid => 'variance_p_int2', aggtransfn => 'int2_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'variance_p_int2' }, { aggfnoid => 'variance(int2)', aggtransfn => 'int2_accum', aggfinalfn => 'numeric_poly_var_samp', aggcombinefn => 'numeric_poly_combine', aggserialfn => 'numeric_poly_serialize', aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int2_accum', aggminvtransfn => 'int2_accum_inv', aggmfinalfn => 'numeric_poly_var_samp', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'variance_p_int2' }, +{ aggfnoid => 'variance_p_float4', aggtransfn => 'float4_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'variance_p_float4' }, { aggfnoid => 'variance(float4)', aggtransfn => 'float4_accum', aggfinalfn => 'float8_var_samp', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'variance_p_float4' }, +{ aggfnoid => 'variance_p_float8', aggtransfn => 'float8_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'variance_p_float8' }, { aggfnoid => 'variance(float8)', aggtransfn => 'float8_accum', aggfinalfn => 'float8_var_samp', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'variance_p_float8' }, +{ aggfnoid => 'variance_p_numeric', aggtransfn => 'numeric_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'variance_p_numeric' }, { aggfnoid => 'variance(numeric)', aggtransfn => 'numeric_accum', aggfinalfn => 'numeric_var_samp', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'numeric_accum', aggminvtransfn => 'numeric_accum_inv', aggmfinalfn => 'numeric_var_samp', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'variance_p_numeric' }, # stddev_pop +{ aggfnoid => 'stddev_pop_p_int8', aggtransfn => 'int8_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'stddev_pop_p_int8' }, { aggfnoid => 'stddev_pop(int8)', aggtransfn => 'int8_accum', aggfinalfn => 'numeric_stddev_pop', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'int8_accum', aggminvtransfn => 'int8_accum_inv', aggmfinalfn => 'numeric_stddev_pop', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'stddev_pop_p_int8' }, +{ aggfnoid => 'stddev_pop_p_int4', aggtransfn => 'int4_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'stddev_pop_p_int4' }, { aggfnoid => 'stddev_pop(int4)', aggtransfn => 'int4_accum', aggfinalfn => 'numeric_poly_stddev_pop', aggcombinefn => 'numeric_poly_combine', @@ -353,7 +576,14 @@ aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int4_accum', aggminvtransfn => 'int4_accum_inv', aggmfinalfn => 'numeric_poly_stddev_pop', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'stddev_pop_p_int4' }, +{ aggfnoid => 'stddev_pop_p_int2', aggtransfn => 'int2_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'stddev_pop_p_int2' }, { aggfnoid => 'stddev_pop(int2)', aggtransfn => 'int2_accum', aggfinalfn => 'numeric_poly_stddev_pop', aggcombinefn => 'numeric_poly_combine', @@ -361,29 +591,58 @@ aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int2_accum', aggminvtransfn => 'int2_accum_inv', aggmfinalfn => 'numeric_poly_stddev_pop', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'stddev_pop_p_int2' }, +{ aggfnoid => 'stddev_pop_p_float4', aggtransfn => 'float4_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'stddev_pop_p_float4' }, { aggfnoid => 'stddev_pop(float4)', aggtransfn => 'float4_accum', aggfinalfn => 'float8_stddev_pop', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'stddev_pop_p_float4' }, +{ aggfnoid => 'stddev_pop_p_float8', aggtransfn => 'float8_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'stddev_pop_p_float8' }, { aggfnoid => 'stddev_pop(float8)', aggtransfn => 'float8_accum', aggfinalfn => 'float8_stddev_pop', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'stddev_pop_p_float8' }, +{ aggfnoid => 'stddev_pop_p_numeric', aggtransfn => 'numeric_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'stddev_pop_p_numeric' }, { aggfnoid => 'stddev_pop(numeric)', aggtransfn => 'numeric_accum', aggfinalfn => 'numeric_stddev_pop', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'numeric_accum', aggminvtransfn => 'numeric_accum_inv', aggmfinalfn => 'numeric_stddev_pop', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'stddev_pop_p_numeric' }, # stddev_samp +{ aggfnoid => 'stddev_samp_p_int8', aggtransfn => 'int8_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'stddev_samp_p_int8' }, { aggfnoid => 'stddev_samp(int8)', aggtransfn => 'int8_accum', aggfinalfn => 'numeric_stddev_samp', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'int8_accum', aggminvtransfn => 'int8_accum_inv', aggmfinalfn => 'numeric_stddev_samp', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'stddev_samp_p_int8' }, +{ aggfnoid => 'stddev_samp_p_int4', aggtransfn => 'int4_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'stddev_samp_p_int4' }, { aggfnoid => 'stddev_samp(int4)', aggtransfn => 'int4_accum', aggfinalfn => 'numeric_poly_stddev_samp', aggcombinefn => 'numeric_poly_combine', @@ -391,7 +650,14 @@ aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int4_accum', aggminvtransfn => 'int4_accum_inv', aggmfinalfn => 'numeric_poly_stddev_samp', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'stddev_samp_p_int4' }, +{ aggfnoid => 'stddev_samp_p_int2', aggtransfn => 'int2_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'stddev_samp_p_int2' }, { aggfnoid => 'stddev_samp(int2)', aggtransfn => 'int2_accum', aggfinalfn => 'numeric_poly_stddev_samp', aggcombinefn => 'numeric_poly_combine', @@ -399,29 +665,58 @@ aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int2_accum', aggminvtransfn => 'int2_accum_inv', aggmfinalfn => 'numeric_poly_stddev_samp', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'stddev_samp_p_int2' }, +{ aggfnoid => 'stddev_samp_p_float4', aggtransfn => 'float4_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'stddev_samp_p_float4' }, { aggfnoid => 'stddev_samp(float4)', aggtransfn => 'float4_accum', aggfinalfn => 'float8_stddev_samp', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'stddev_samp_p_float4' }, +{ aggfnoid => 'stddev_samp_p_float8', aggtransfn => 'float8_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'stddev_samp_p_float8' }, { aggfnoid => 'stddev_samp(float8)', aggtransfn => 'float8_accum', aggfinalfn => 'float8_stddev_samp', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'stddev_samp_p_float8' }, +{ aggfnoid => 'stddev_samp_p_numeric', aggtransfn => 'numeric_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'stddev_samp_p_numeric' }, { aggfnoid => 'stddev_samp(numeric)', aggtransfn => 'numeric_accum', aggfinalfn => 'numeric_stddev_samp', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'numeric_accum', aggminvtransfn => 'numeric_accum_inv', aggmfinalfn => 'numeric_stddev_samp', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'stddev_samp_p_numeric' }, # stddev: historical Postgres syntax for stddev_samp +{ aggfnoid => 'stddev_p_int8', aggtransfn => 'int8_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'stddev_p_int8' }, { aggfnoid => 'stddev(int8)', aggtransfn => 'int8_accum', aggfinalfn => 'numeric_stddev_samp', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'int8_accum', aggminvtransfn => 'int8_accum_inv', aggmfinalfn => 'numeric_stddev_samp', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'stddev_p_int8' }, +{ aggfnoid => 'stddev_p_int4', aggtransfn => 'int4_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'stddev_p_int4' }, { aggfnoid => 'stddev(int4)', aggtransfn => 'int4_accum', aggfinalfn => 'numeric_poly_stddev_samp', aggcombinefn => 'numeric_poly_combine', @@ -429,7 +724,14 @@ aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int4_accum', aggminvtransfn => 'int4_accum_inv', aggmfinalfn => 'numeric_poly_stddev_samp', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'stddev_p_int4' }, +{ aggfnoid => 'stddev_p_int2', aggtransfn => 'int2_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'stddev_p_int2' }, { aggfnoid => 'stddev(int2)', aggtransfn => 'int2_accum', aggfinalfn => 'numeric_poly_stddev_samp', aggcombinefn => 'numeric_poly_combine', @@ -437,138 +739,253 @@ aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int2_accum', aggminvtransfn => 'int2_accum_inv', aggmfinalfn => 'numeric_poly_stddev_samp', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'stddev_p_int2' }, +{ aggfnoid => 'stddev_p_float4', aggtransfn => 'float4_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'stddev_p_float4' }, { aggfnoid => 'stddev(float4)', aggtransfn => 'float4_accum', aggfinalfn => 'float8_stddev_samp', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'stddev_p_float4' }, +{ aggfnoid => 'stddev_p_float8', aggtransfn => 'float8_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'stddev_p_float8' }, { aggfnoid => 'stddev(float8)', aggtransfn => 'float8_accum', aggfinalfn => 'float8_stddev_samp', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'stddev_p_float8' }, +{ aggfnoid => 'stddev_p_numeric', aggtransfn => 'numeric_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'stddev_p_numeric' }, { aggfnoid => 'stddev(numeric)', aggtransfn => 'numeric_accum', aggfinalfn => 'numeric_stddev_samp', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'numeric_accum', aggminvtransfn => 'numeric_accum_inv', aggmfinalfn => 'numeric_stddev_samp', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'stddev_p_numeric' }, # SQL2003 binary regression aggregates { aggfnoid => 'regr_count', aggtransfn => 'int8inc_float8_float8', - aggcombinefn => 'int8pl', aggtranstype => 'int8', agginitval => '0' }, + aggcombinefn => 'int8pl', aggtranstype => 'int8', agginitval => '0', + aggpartialfn => 'regr_count' }, +{ aggfnoid => 'regr_sxx_p', aggtransfn => 'float8_regr_accum', + aggcombinefn => 'float8_regr_combine', aggtranstype => '_float8', + agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_sxx_p' }, { aggfnoid => 'regr_sxx', aggtransfn => 'float8_regr_accum', aggfinalfn => 'float8_regr_sxx', aggcombinefn => 'float8_regr_combine', - aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_sxx_p' }, +{ aggfnoid => 'regr_syy_p', aggtransfn => 'float8_regr_accum', + aggcombinefn => 'float8_regr_combine', aggtranstype => '_float8', + agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_syy_p' }, { aggfnoid => 'regr_syy', aggtransfn => 'float8_regr_accum', aggfinalfn => 'float8_regr_syy', aggcombinefn => 'float8_regr_combine', - aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_syy_p' }, +{ aggfnoid => 'regr_sxy_p', aggtransfn => 'float8_regr_accum', + aggcombinefn => 'float8_regr_combine', aggtranstype => '_float8', + agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_sxy_p' }, { aggfnoid => 'regr_sxy', aggtransfn => 'float8_regr_accum', aggfinalfn => 'float8_regr_sxy', aggcombinefn => 'float8_regr_combine', - aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_sxy_p' }, +{ aggfnoid => 'regr_avgx_p', aggtransfn => 'float8_regr_accum', + aggcombinefn => 'float8_regr_combine', aggtranstype => '_float8', + agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_avgx_p' }, { aggfnoid => 'regr_avgx', aggtransfn => 'float8_regr_accum', aggfinalfn => 'float8_regr_avgx', aggcombinefn => 'float8_regr_combine', - aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_avgx_p' }, +{ aggfnoid => 'regr_avgy_p', aggtransfn => 'float8_regr_accum', + aggcombinefn => 'float8_regr_combine', aggtranstype => '_float8', + agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_avgy_p' }, { aggfnoid => 'regr_avgy', aggtransfn => 'float8_regr_accum', aggfinalfn => 'float8_regr_avgy', aggcombinefn => 'float8_regr_combine', - aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_avgy_p' }, +{ aggfnoid => 'regr_r2_p', aggtransfn => 'float8_regr_accum', + aggcombinefn => 'float8_regr_combine', aggtranstype => '_float8', + agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_r2_p' }, { aggfnoid => 'regr_r2', aggtransfn => 'float8_regr_accum', aggfinalfn => 'float8_regr_r2', aggcombinefn => 'float8_regr_combine', - aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_r2_p' }, +{ aggfnoid => 'regr_slope_p', aggtransfn => 'float8_regr_accum', + aggcombinefn => 'float8_regr_combine', aggtranstype => '_float8', + agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_slope_p' }, { aggfnoid => 'regr_slope', aggtransfn => 'float8_regr_accum', aggfinalfn => 'float8_regr_slope', aggcombinefn => 'float8_regr_combine', - aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_slope_p' }, +{ aggfnoid => 'regr_intercept_p', aggtransfn => 'float8_regr_accum', + aggcombinefn => 'float8_regr_combine', aggtranstype => '_float8', + agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_intercept_p' }, { aggfnoid => 'regr_intercept', aggtransfn => 'float8_regr_accum', aggfinalfn => 'float8_regr_intercept', aggcombinefn => 'float8_regr_combine', - aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_intercept_p' }, +{ aggfnoid => 'covar_pop_p', aggtransfn => 'float8_regr_accum', + aggcombinefn => 'float8_regr_combine', aggtranstype => '_float8', + agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'covar_pop_p' }, { aggfnoid => 'covar_pop', aggtransfn => 'float8_regr_accum', aggfinalfn => 'float8_covar_pop', aggcombinefn => 'float8_regr_combine', - aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'covar_pop_p' }, +{ aggfnoid => 'covar_samp_p', aggtransfn => 'float8_regr_accum', + aggcombinefn => 'float8_regr_combine', aggtranstype => '_float8', + agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'covar_samp_p' }, { aggfnoid => 'covar_samp', aggtransfn => 'float8_regr_accum', aggfinalfn => 'float8_covar_samp', aggcombinefn => 'float8_regr_combine', - aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'covar_samp_p' }, +{ aggfnoid => 'corr_p', aggtransfn => 'float8_regr_accum', + aggcombinefn => 'float8_regr_combine', aggtranstype => '_float8', + agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'corr_p' }, { aggfnoid => 'corr', aggtransfn => 'float8_regr_accum', aggfinalfn => 'float8_corr', aggcombinefn => 'float8_regr_combine', - aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'corr_p' }, # boolean-and and boolean-or { aggfnoid => 'bool_and', aggtransfn => 'booland_statefunc', aggcombinefn => 'booland_statefunc', aggmtransfn => 'bool_accum', aggminvtransfn => 'bool_accum_inv', aggmfinalfn => 'bool_alltrue', aggsortop => '<(bool,bool)', aggtranstype => 'bool', - aggmtranstype => 'internal', aggmtransspace => '16' }, + aggmtranstype => 'internal', aggmtransspace => '16', + aggpartialfn => 'bool_and' }, { aggfnoid => 'bool_or', aggtransfn => 'boolor_statefunc', aggcombinefn => 'boolor_statefunc', aggmtransfn => 'bool_accum', aggminvtransfn => 'bool_accum_inv', aggmfinalfn => 'bool_anytrue', aggsortop => '>(bool,bool)', aggtranstype => 'bool', - aggmtranstype => 'internal', aggmtransspace => '16' }, + aggmtranstype => 'internal', aggmtransspace => '16', + aggpartialfn => 'bool_or' }, { aggfnoid => 'every', aggtransfn => 'booland_statefunc', aggcombinefn => 'booland_statefunc', aggmtransfn => 'bool_accum', aggminvtransfn => 'bool_accum_inv', aggmfinalfn => 'bool_alltrue', aggsortop => '<(bool,bool)', aggtranstype => 'bool', - aggmtranstype => 'internal', aggmtransspace => '16' }, + aggmtranstype => 'internal', aggmtransspace => '16', + aggpartialfn => 'every' }, # bitwise integer { aggfnoid => 'bit_and(int2)', aggtransfn => 'int2and', - aggcombinefn => 'int2and', aggtranstype => 'int2' }, + aggcombinefn => 'int2and', aggtranstype => 'int2', + aggpartialfn => 'bit_and(int2)' }, { aggfnoid => 'bit_or(int2)', aggtransfn => 'int2or', aggcombinefn => 'int2or', - aggtranstype => 'int2' }, + aggtranstype => 'int2', + aggpartialfn => 'bit_or(int2)' }, { aggfnoid => 'bit_xor(int2)', aggtransfn => 'int2xor', - aggcombinefn => 'int2xor', aggtranstype => 'int2' }, + aggcombinefn => 'int2xor', aggtranstype => 'int2', + aggpartialfn => 'bit_xor(int2)' }, { aggfnoid => 'bit_and(int4)', aggtransfn => 'int4and', - aggcombinefn => 'int4and', aggtranstype => 'int4' }, + aggcombinefn => 'int4and', aggtranstype => 'int4', + aggpartialfn => 'bit_and(int4)' }, { aggfnoid => 'bit_or(int4)', aggtransfn => 'int4or', aggcombinefn => 'int4or', - aggtranstype => 'int4' }, + aggtranstype => 'int4', + aggpartialfn => 'bit_or(int4)' }, { aggfnoid => 'bit_xor(int4)', aggtransfn => 'int4xor', - aggcombinefn => 'int4xor', aggtranstype => 'int4' }, + aggcombinefn => 'int4xor', aggtranstype => 'int4', + aggpartialfn => 'bit_xor(int4)' }, { aggfnoid => 'bit_and(int8)', aggtransfn => 'int8and', - aggcombinefn => 'int8and', aggtranstype => 'int8' }, + aggcombinefn => 'int8and', aggtranstype => 'int8', + aggpartialfn => 'bit_and(int8)' }, { aggfnoid => 'bit_or(int8)', aggtransfn => 'int8or', aggcombinefn => 'int8or', - aggtranstype => 'int8' }, + aggtranstype => 'int8', + aggpartialfn => 'bit_or(int8)' }, { aggfnoid => 'bit_xor(int8)', aggtransfn => 'int8xor', - aggcombinefn => 'int8xor', aggtranstype => 'int8' }, + aggcombinefn => 'int8xor', aggtranstype => 'int8', + aggpartialfn => 'bit_xor(int8)' }, { aggfnoid => 'bit_and(bit)', aggtransfn => 'bitand', aggcombinefn => 'bitand', - aggtranstype => 'bit' }, + aggtranstype => 'bit', + aggpartialfn => 'bit_and(bit)' }, { aggfnoid => 'bit_or(bit)', aggtransfn => 'bitor', aggcombinefn => 'bitor', - aggtranstype => 'bit' }, + aggtranstype => 'bit', + aggpartialfn => 'bit_or(bit)' }, { aggfnoid => 'bit_xor(bit)', aggtransfn => 'bitxor', aggcombinefn => 'bitxor', - aggtranstype => 'bit' }, + aggtranstype => 'bit', + aggpartialfn => 'bit_xor(bit)' }, # xml { aggfnoid => 'xmlagg', aggtransfn => 'xmlconcat2', aggtranstype => 'xml' }, # array +{ aggfnoid => 'array_agg_p_anynonarray', aggtransfn => 'array_agg_transfn', + aggfinalfn => 'array_agg_serialize', aggcombinefn => 'array_agg_combine', + aggserialfn => 'array_agg_serialize', aggdeserialfn => 'array_agg_deserialize', + aggtranstype => 'internal', + aggpartialfn => 'array_agg_p_anynonarray' }, { aggfnoid => 'array_agg(anynonarray)', aggtransfn => 'array_agg_transfn', aggfinalfn => 'array_agg_finalfn', aggcombinefn => 'array_agg_combine', aggserialfn => 'array_agg_serialize', aggdeserialfn => 'array_agg_deserialize', aggfinalextra => 't', - aggtranstype => 'internal' }, + aggtranstype => 'internal', + aggpartialfn => 'array_agg_p_anynonarray' }, +{ aggfnoid => 'array_agg_p_anyarray', aggtransfn => 'array_agg_array_transfn', + aggfinalfn => 'array_agg_array_serialize', + aggcombinefn => 'array_agg_array_combine', + aggserialfn => 'array_agg_array_serialize', + aggdeserialfn => 'array_agg_array_deserialize', aggtranstype => 'internal', + aggpartialfn => 'array_agg_p_anyarray' }, { aggfnoid => 'array_agg(anyarray)', aggtransfn => 'array_agg_array_transfn', aggfinalfn => 'array_agg_array_finalfn', aggcombinefn => 'array_agg_array_combine', aggserialfn => 'array_agg_array_serialize', aggdeserialfn => 'array_agg_array_deserialize', aggfinalextra => 't', - aggtranstype => 'internal' }, + aggtranstype => 'internal', + aggpartialfn => 'array_agg_p_anyarray' }, # text +{ aggfnoid => 'string_agg_p_text_text', aggtransfn => 'string_agg_transfn', + aggfinalfn => 'string_agg_serialize', aggcombinefn => 'string_agg_combine', + aggserialfn => 'string_agg_serialize', + aggdeserialfn => 'string_agg_deserialize', aggtranstype => 'internal', + aggpartialfn => 'string_agg_p_text_text' }, { aggfnoid => 'string_agg(text,text)', aggtransfn => 'string_agg_transfn', aggfinalfn => 'string_agg_finalfn', aggcombinefn => 'string_agg_combine', aggserialfn => 'string_agg_serialize', - aggdeserialfn => 'string_agg_deserialize', aggtranstype => 'internal' }, + aggdeserialfn => 'string_agg_deserialize', aggtranstype => 'internal', + aggpartialfn => 'string_agg_p_text_text' }, # bytea +{ aggfnoid => 'string_agg_p_bytea_bytea', + aggtransfn => 'bytea_string_agg_transfn', aggfinalfn => 'string_agg_serialize', + aggcombinefn => 'string_agg_combine', aggserialfn => 'string_agg_serialize', + aggdeserialfn => 'string_agg_deserialize', aggtranstype => 'internal', + aggpartialfn => 'string_agg_p_bytea_bytea' }, { aggfnoid => 'string_agg(bytea,bytea)', aggtransfn => 'bytea_string_agg_transfn', aggfinalfn => 'bytea_string_agg_finalfn', aggcombinefn => 'string_agg_combine', aggserialfn => 'string_agg_serialize', - aggdeserialfn => 'string_agg_deserialize', aggtranstype => 'internal' }, + aggdeserialfn => 'string_agg_deserialize', aggtranstype => 'internal', + aggpartialfn => 'string_agg_p_bytea_bytea' }, # range { aggfnoid => 'range_intersect_agg(anyrange)', aggtransfn => 'range_intersect_agg_transfn', - aggcombinefn => 'range_intersect_agg_transfn', aggtranstype => 'anyrange' }, + aggcombinefn => 'range_intersect_agg_transfn', aggtranstype => 'anyrange', + aggpartialfn => 'range_intersect_agg(anyrange)' }, { aggfnoid => 'range_intersect_agg(anymultirange)', aggtransfn => 'multirange_intersect_agg_transfn', aggcombinefn => 'multirange_intersect_agg_transfn', - aggtranstype => 'anymultirange' }, + aggtranstype => 'anymultirange', + aggpartialfn => 'range_intersect_agg(anymultirange)' }, { aggfnoid => 'range_agg(anyrange)', aggtransfn => 'range_agg_transfn', aggfinalfn => 'range_agg_finalfn', aggfinalextra => 't', aggtranstype => 'internal' }, @@ -658,6 +1075,7 @@ # any_value { aggfnoid => 'any_value(anyelement)', aggtransfn => 'any_value_transfn', - aggcombinefn => 'any_value_transfn', aggtranstype => 'anyelement' }, + aggcombinefn => 'any_value_transfn', aggtranstype => 'anyelement', + aggpartialfn => 'any_value(anyelement)' }, ] diff --git a/src/include/catalog/pg_aggregate.h b/src/include/catalog/pg_aggregate.h index 3112881193..b8fb0372b3 100644 --- a/src/include/catalog/pg_aggregate.h +++ b/src/include/catalog/pg_aggregate.h @@ -55,6 +55,9 @@ CATALOG(pg_aggregate,2600,AggregateRelationId) /* function to convert bytea to transtype (0 if none) */ regproc aggdeserialfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc); + /* special aggregate function for partial aggregation (0 if none) */ + regproc aggpartialfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc); + /* forward function for moving-aggregate mode (0 if none) */ regproc aggmtransfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc); @@ -164,6 +167,7 @@ extern ObjectAddress AggregateCreate(const char *aggName, List *aggmtransfnName, List *aggminvtransfnName, List *aggmfinalfnName, + List *aggpartialfnName, bool finalfnExtraArgs, bool mfinalfnExtraArgs, char finalfnModify, diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat index 6996073989..1c4087ad8c 100644 --- a/src/include/catalog/pg_proc.dat +++ b/src/include/catalog/pg_proc.dat @@ -1658,6 +1658,10 @@ { oid => '2334', descr => 'aggregate final function', proname => 'array_agg_finalfn', proisstrict => 'f', prorettype => 'anyarray', proargtypes => 'internal anynonarray', prosrc => 'array_agg_finalfn' }, +{ oid => '6', descr => 'partial aggregation for array_agg(anynonarray)', + proname => 'array_agg_p_anynonarray', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'anynonarray', + prosrc => 'aggregate_dummy' }, { oid => '2335', descr => 'concatenate aggregate input into an array', proname => 'array_agg', prokind => 'a', proisstrict => 'f', prorettype => 'anyarray', proargtypes => 'anynonarray', @@ -1680,6 +1684,9 @@ proname => 'array_agg_array_finalfn', proisstrict => 'f', prorettype => 'anyarray', proargtypes => 'internal anyarray', prosrc => 'array_agg_array_finalfn' }, +{ oid => '7', descr => 'partial aggregation for array_agg(anyarray)', + proname => 'array_agg_p_anyarray', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'anyarray', prosrc => 'aggregate_dummy' }, { oid => '4053', descr => 'concatenate aggregate input into an array', proname => 'array_agg', prokind => 'a', proisstrict => 'f', prorettype => 'anyarray', proargtypes => 'anyarray', @@ -4989,6 +4996,10 @@ { oid => '3536', descr => 'aggregate final function', proname => 'string_agg_finalfn', proisstrict => 'f', prorettype => 'text', proargtypes => 'internal', prosrc => 'string_agg_finalfn' }, +{ oid => '8', descr => 'partial aggregation for string_agg(text, text)', + proname => 'string_agg_p_text_text', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'text text', + prosrc => 'aggregate_dummy' }, { oid => '3538', descr => 'concatenate aggregate input into a string', proname => 'string_agg', prokind => 'a', proisstrict => 'f', prorettype => 'text', proargtypes => 'text text', @@ -5001,6 +5012,10 @@ proname => 'bytea_string_agg_finalfn', proisstrict => 'f', prorettype => 'bytea', proargtypes => 'internal', prosrc => 'bytea_string_agg_finalfn' }, +{ oid => '9', descr => 'partial aggregation for string_agg(bytea, bytea)', + proname => 'string_agg_p_bytea_bytea', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'bytea bytea', + prosrc => 'aggregate_dummy' }, { oid => '3545', descr => 'concatenate aggregate input into a bytea', proname => 'string_agg', prokind => 'a', proisstrict => 'f', prorettype => 'bytea', proargtypes => 'bytea bytea', @@ -6579,36 +6594,61 @@ # Aggregates (moved here from pg_aggregate for 7.3) +{ oid => '111', descr => 'partial aggregation for avg(int8)', + proname => 'avg_p_int8', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, { oid => '2100', descr => 'the average (arithmetic mean) as numeric of all bigint values', proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, +{ oid => '226', descr => 'partial aggregation for avg(int4)', + proname => 'avg_p_int4', prokind => 'a', proisstrict => 'f', + prorettype => '_int8', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, { oid => '2101', descr => 'the average (arithmetic mean) as numeric of all integer values', proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, +{ oid => '388', descr => 'partial aggregation for avg(int2)', + proname => 'avg_p_int2', prokind => 'a', proisstrict => 'f', + prorettype => '_int8', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, { oid => '2102', descr => 'the average (arithmetic mean) as numeric of all smallint values', proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, +{ oid => '389', descr => 'partial aggregation for avg(numeric)', + proname => 'avg_p_numeric', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, { oid => '2103', descr => 'the average (arithmetic mean) as numeric of all numeric values', proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, +{ oid => '560', descr => 'partial aggregation for avg(float4)', + proname => 'avg_p_float4', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, { oid => '2104', descr => 'the average (arithmetic mean) as float8 of all float4 values', proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, +{ oid => '561', descr => 'partial aggregation for avg(float8)', + proname => 'avg_p_float8', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, { oid => '2105', descr => 'the average (arithmetic mean) as float8 of all float8 values', proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, +{ oid => '562', descr => 'partial aggregation for avg(interval)', + proname => 'avg_p_interval', prokind => 'a', proisstrict => 'f', + prorettype => '_interval', proargtypes => 'interval', + prosrc => 'aggregate_dummy' }, { oid => '2106', descr => 'the average (arithmetic mean) as interval of all interval values', proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'interval', proargtypes => 'interval', prosrc => 'aggregate_dummy' }, +{ oid => '563', descr => 'partial aggregation for sum(int8)', + proname => 'sum_p_int8', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, { oid => '2107', descr => 'sum as numeric across all bigint input values', proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, @@ -6631,6 +6671,9 @@ proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'interval', proargtypes => 'interval', prosrc => 'aggregate_dummy' }, +{ oid => '564', descr => 'partial aggregation for sum(numeric)', + proname => 'sum_p_numeric', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, { oid => '2114', descr => 'sum as numeric across all numeric input values', proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, @@ -6789,153 +6832,261 @@ proname => 'int8inc_support', prorettype => 'internal', proargtypes => 'internal', prosrc => 'int8inc_support' }, +{ oid => '565', descr => 'partial aggregation for var_pop(int8)', + proname => 'var_pop_p_int8', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, { oid => '2718', descr => 'population variance of bigint input values (square of the population standard deviation)', proname => 'var_pop', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, +{ oid => '566', descr => 'partial aggregation for var_pop(int4)', + proname => 'var_pop_p_int4', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, { oid => '2719', descr => 'population variance of integer input values (square of the population standard deviation)', proname => 'var_pop', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, +{ oid => '567', descr => 'partial aggregation for var_pop(int2)', + proname => 'var_pop_p_int2', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, { oid => '2720', descr => 'population variance of smallint input values (square of the population standard deviation)', proname => 'var_pop', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, +{ oid => '568', descr => 'partial aggregation for var_pop(float4)', + proname => 'var_pop_p_float4', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, { oid => '2721', descr => 'population variance of float4 input values (square of the population standard deviation)', proname => 'var_pop', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, +{ oid => '569', descr => 'partial aggregation for var_pop(float8)', + proname => 'var_pop_p_float8', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, { oid => '2722', descr => 'population variance of float8 input values (square of the population standard deviation)', proname => 'var_pop', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, +{ oid => '570', descr => 'partial aggregation for var_pop(numeric)', + proname => 'var_pop_p_numeric', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, { oid => '2723', descr => 'population variance of numeric input values (square of the population standard deviation)', proname => 'var_pop', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, +{ oid => '571', descr => 'partial aggregation for var_samp(int8)', + proname => 'var_samp_p_int8', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, { oid => '2641', descr => 'sample variance of bigint input values (square of the sample standard deviation)', proname => 'var_samp', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, +{ oid => '572', descr => 'partial aggregation for var_samp(int4)', + proname => 'var_samp_p_int4', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, { oid => '2642', descr => 'sample variance of integer input values (square of the sample standard deviation)', proname => 'var_samp', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, +{ oid => '573', descr => 'partial aggregation for var_samp(int2)', + proname => 'var_samp_p_int2', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, { oid => '2643', descr => 'sample variance of smallint input values (square of the sample standard deviation)', proname => 'var_samp', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, +{ oid => '574', descr => 'partial aggregation for var_samp(float4)', + proname => 'var_samp_p_float4', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, { oid => '2644', descr => 'sample variance of float4 input values (square of the sample standard deviation)', proname => 'var_samp', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, +{ oid => '575', descr => 'partial aggregation for var_samp(float8)', + proname => 'var_samp_p_float8', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, { oid => '2645', descr => 'sample variance of float8 input values (square of the sample standard deviation)', proname => 'var_samp', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, +{ oid => '576', descr => 'partial aggregation for var_samp(numeric)', + proname => 'var_samp_p_numeric', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, { oid => '2646', descr => 'sample variance of numeric input values (square of the sample standard deviation)', proname => 'var_samp', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, +{ oid => '577', descr => 'partial aggregation for variance(int8)', + proname => 'variance_p_int8', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, { oid => '2148', descr => 'historical alias for var_samp', proname => 'variance', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, +{ oid => '578', descr => 'partial aggregation for variance(int4)', + proname => 'variance_p_int4', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, { oid => '2149', descr => 'historical alias for var_samp', proname => 'variance', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, +{ oid => '579', descr => 'partial aggregation for variance(int2)', + proname => 'variance_p_int2', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, { oid => '2150', descr => 'historical alias for var_samp', proname => 'variance', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, +{ oid => '580', descr => 'partial aggregation for variance(float4)', + proname => 'variance_p_float4', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, { oid => '2151', descr => 'historical alias for var_samp', proname => 'variance', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, +{ oid => '581', descr => 'partial aggregation for variance(float8)', + proname => 'variance_p_float8', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, { oid => '2152', descr => 'historical alias for var_samp', proname => 'variance', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, +{ oid => '582', descr => 'partial aggregation for variance(numeric)', + proname => 'variance_p_numeric', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, { oid => '2153', descr => 'historical alias for var_samp', proname => 'variance', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, +{ oid => '583', descr => 'partial aggregation for stddev_pop(int8)', + proname => 'stddev_pop_p_int8', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, { oid => '2724', descr => 'population standard deviation of bigint input values', proname => 'stddev_pop', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, +{ oid => '703', descr => 'partial aggregation for stddev_pop(int4)', + proname => 'stddev_pop_p_int4', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, { oid => '2725', descr => 'population standard deviation of integer input values', proname => 'stddev_pop', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, +{ oid => '726', descr => 'partial aggregation for stddev_pop(int2)', + proname => 'stddev_pop_p_int2', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, { oid => '2726', descr => 'population standard deviation of smallint input values', proname => 'stddev_pop', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, +{ oid => '786', descr => 'partial aggregation for stddev_pop(float4)', + proname => 'stddev_pop_p_float4', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, { oid => '2727', descr => 'population standard deviation of float4 input values', proname => 'stddev_pop', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, +{ oid => '787', descr => 'partial aggregation for stddev_pop(float8)', + proname => 'stddev_pop_p_float8', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, { oid => '2728', descr => 'population standard deviation of float8 input values', proname => 'stddev_pop', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, +{ oid => '788', descr => 'partial aggregation for stddev_pop(numeric)', + proname => 'stddev_pop_p_numeric', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, { oid => '2729', descr => 'population standard deviation of numeric input values', proname => 'stddev_pop', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, +{ oid => '789', descr => 'partial aggregation for stddev_samp(int8)', + proname => 'stddev_samp_p_int8', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, { oid => '2712', descr => 'sample standard deviation of bigint input values', proname => 'stddev_samp', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, +{ oid => '811', descr => 'partial aggregation for stddev_samp(int4)', + proname => 'stddev_samp_p_int4', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, { oid => '2713', descr => 'sample standard deviation of integer input values', proname => 'stddev_samp', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, +{ oid => '812', descr => 'partial aggregation for stddev_samp(int2)', + proname => 'stddev_samp_p_int2', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, { oid => '2714', descr => 'sample standard deviation of smallint input values', proname => 'stddev_samp', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, +{ oid => '813', descr => 'partial aggregation for stddev_samp(float4)', + proname => 'stddev_samp_p_float4', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, { oid => '2715', descr => 'sample standard deviation of float4 input values', proname => 'stddev_samp', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, +{ oid => '814', descr => 'partial aggregation for stddev_samp(float8)', + proname => 'stddev_samp_p_float8', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, { oid => '2716', descr => 'sample standard deviation of float8 input values', proname => 'stddev_samp', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, +{ oid => '815', descr => 'partial aggregation for stddev_samp(numeric)', + proname => 'stddev_samp_p_numeric', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, { oid => '2717', descr => 'sample standard deviation of numeric input values', proname => 'stddev_samp', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, +{ oid => '816', descr => 'partial aggregation for stddev(int8)', + proname => 'stddev_p_int8', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, { oid => '2154', descr => 'historical alias for stddev_samp', proname => 'stddev', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, +{ oid => '970', descr => 'partial aggregation for stddev(int4)', + proname => 'stddev_p_int4', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, { oid => '2155', descr => 'historical alias for stddev_samp', proname => 'stddev', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, +{ oid => '1382', descr => 'partial aggregation for stddev(int2)', + proname => 'stddev_p_int2', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, { oid => '2156', descr => 'historical alias for stddev_samp', proname => 'stddev', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, +{ oid => '1524', descr => 'partial aggregation for stddev(float4)', + proname => 'stddev_p_float4', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, { oid => '2157', descr => 'historical alias for stddev_samp', proname => 'stddev', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, +{ oid => '1533', descr => 'partial aggregation for stddev(float8)', + proname => 'stddev_p_float8', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, { oid => '2158', descr => 'historical alias for stddev_samp', proname => 'stddev', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, +{ oid => '1566', descr => 'partial aggregation for stddev(numeric)', + proname => 'stddev_p_numeric', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, { oid => '2159', descr => 'historical alias for stddev_samp', proname => 'stddev', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'numeric', @@ -6946,52 +7097,97 @@ proname => 'regr_count', prokind => 'a', proisstrict => 'f', prorettype => 'int8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, +{ oid => '1568', descr => 'partial aggregation for regr_sxx(float8, float8)', + proname => 'regr_sxx_p', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8 float8', + prosrc => 'aggregate_dummy' }, { oid => '2819', descr => 'sum of squares of the independent variable (sum(X^2) - sum(X)^2/N)', proname => 'regr_sxx', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, +{ oid => '2173', descr => 'partial aggregation for regr_syy(float8, float8)', + proname => 'regr_syy_p', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8 float8', + prosrc => 'aggregate_dummy' }, { oid => '2820', descr => 'sum of squares of the dependent variable (sum(Y^2) - sum(Y)^2/N)', proname => 'regr_syy', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, +{ oid => '3813', descr => 'partial aggregation for regr_sxy(float8, float8)', + proname => 'regr_sxy_p', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8 float8', + prosrc => 'aggregate_dummy' }, { oid => '2821', descr => 'sum of products of independent times dependent variable (sum(X*Y) - sum(X) * sum(Y)/N)', proname => 'regr_sxy', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, +{ oid => '3814', descr => 'partial aggregation for regr_avgx(float8, float8)', + proname => 'regr_avgx_p', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8 float8', + prosrc => 'aggregate_dummy' }, { oid => '2822', descr => 'average of the independent variable (sum(X)/N)', proname => 'regr_avgx', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, +{ oid => '4551', descr => 'partial aggregation for regr_avgy(float8, float8)', + proname => 'regr_avgy_p', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8 float8', + prosrc => 'aggregate_dummy' }, { oid => '2823', descr => 'average of the dependent variable (sum(Y)/N)', proname => 'regr_avgy', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, +{ oid => '4552', descr => 'partial aggregation for regr_r2(float8, float8)', + proname => 'regr_r2_p', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8 float8', + prosrc => 'aggregate_dummy' }, { oid => '2824', descr => 'square of the correlation coefficient', proname => 'regr_r2', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, +{ oid => '4553', descr => 'partial aggregation for regr_slope(float8, float8)', + proname => 'regr_slope_p', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8 float8', + prosrc => 'aggregate_dummy' }, { oid => '2825', descr => 'slope of the least-squares-fit linear equation determined by the (X, Y) pairs', proname => 'regr_slope', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, +{ oid => '4554', + descr => 'partial aggregation for regr_intercept(float8, float8)', + proname => 'regr_intercept_p', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8 float8', + prosrc => 'aggregate_dummy' }, { oid => '2826', descr => 'y-intercept of the least-squares-fit linear equation determined by the (X, Y) pairs', proname => 'regr_intercept', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, +{ oid => '4555', descr => 'partial aggregation for covar_pop(float8, float8)', + proname => 'covar_pop_p', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8 float8', + prosrc => 'aggregate_dummy' }, { oid => '2827', descr => 'population covariance', proname => 'covar_pop', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, +{ oid => '4556', descr => 'partial aggregation for covar_samp(float8, float8)', + proname => 'covar_samp_p', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8 float8', + prosrc => 'aggregate_dummy' }, { oid => '2828', descr => 'sample covariance', proname => 'covar_samp', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, +{ oid => '4557', descr => 'partial aggregation for corr(float8, float8)', + proname => 'corr_p', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8 float8', + prosrc => 'aggregate_dummy' }, { oid => '2829', descr => 'correlation coefficient', proname => 'corr', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, diff --git a/src/test/regress/expected/create_aggregate.out b/src/test/regress/expected/create_aggregate.out index dcf6909423..73e4cfde0f 100644 --- a/src/test/regress/expected/create_aggregate.out +++ b/src/test/regress/expected/create_aggregate.out @@ -322,3 +322,187 @@ WARNING: aggregate attribute "Finalfunc_extra" not recognized WARNING: aggregate attribute "Finalfunc_modify" not recognized WARNING: aggregate attribute "Parallel" not recognized ERROR: aggregate stype must be specified +-- invalid: finalfunc is specified and stype is not internal +-- even though aggpartialfunc is equal to the aggregate name +CREATE AGGREGATE udf_avg_int4_invalid(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + initcond = '{0,0}', + aggpartialfunc = udf_avg_int4_invalid +); +ERROR: udf_avg_int4_invalid is not its own aggpartialfunc +-- invalid: finalfunc is not equal to serialfunc and stype is internal +-- even though aggpartialfunc is equal to the aggregate name +CREATE AGGREGATE udf_avg_int8_invalid(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + combinefunc = int8_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = udf_avg_int8_invalid +); +ERROR: udf_avg_int8_invalid is not its own aggpartialfunc +-- invalid: aggpartialfunc which doesn't exist +CREATE AGGREGATE nonexistent_aggpartial_agg(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_sum, + combinefunc = int8_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = aggpartial_foo +); +ERROR: function aggpartial_foo(bigint) does not exist +-- invalid: aggpartialfunc is specified and combinefunc isn't specified +CREATE AGGREGATE nocombinefunc_agg(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = avg_p_int8 +); +ERROR: aggcombinefnName must be supplied if aggpartialfnName is supplied +-- invalid: combinefunc in agg and combinefunc in aggpartialfunc don't match +CREATE AGGREGATE udf_avg_invalid_combinefunc(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + combinefunc = numeric_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = avg_p_int8 +); +ERROR: combinefunc in aggregate and combinefunc in its aggpartialfunc must match +CREATE FUNCTION udf_float8_accum(_float8, float8) RETURNS _float8 AS +$$ SELECT float8_accum($1, $2) $$ +LANGUAGE SQL; +-- invalid: sfunc in agg and sfunc in aggpartialfunc don't match +CREATE AGGREGATE udf_avg_invalid_sfunc(float8) ( + sfunc = udf_float8_accum, + stype = _float8, + finalfunc = float8_avg, + combinefunc = float8_combine, + initcond = '{0,0,0}', + aggpartialfunc = avg_p_float8 +); +ERROR: sfunc in aggregate and its aggpartialfunc must match +DROP FUNCTION udf_float8_accum(_float8, float8); +-- invalid: initcond in agg and initcond in aggpartialfunc don't match +CREATE AGGREGATE udf_avg_invalid_initcond(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + initcond = '{1,0}', + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + aggpartialfunc = avg_p_int4 +); +ERROR: initcond in aggregate and initcond in its aggpartialfunc must match +-- invalid: aggserialfn in agg and aggserialfn in aggpartialfunc don't match +CREATE AGGREGATE udf_avg_invalid_aggserialfn(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + combinefunc = int8_avg_combine, + serialfunc = numeric_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = avg_p_int8 +); +ERROR: serialfunc in aggregate and serialfunc in its aggpartialfunc must match +-- invalid: aggdeserialfn in agg and aggdeserialfn in aggpartialfunc don't match +CREATE AGGREGATE udf_avg_invalid_aggdeserialfn(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + combinefunc = int8_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = numeric_avg_deserialize, + aggpartialfunc = avg_p_int8 +); +ERROR: deserialfunc in aggregate and deserialfunc in its aggpartialfunc must match +-- invalid: stype is internal and aggpartialfunc's finalfunc +-- isn't serialfunc of agg +CREATE AGGREGATE udf_avg_p_int8(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_avg_serialize, + combinefunc = int8_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize +); +CREATE AGGREGATE udf_avg(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + combinefunc = int8_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = udf_avg_p_int8 +); +ERROR: finalfunc of aggpartialfunc must match serialfunc of aggregate when stype is internal +-- invalid: stype is not internal and aggpartialfunc has finalfunc +CREATE AGGREGATE udf_avg_p_int4_hasfinal(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + initcond = '{0,0}' +); +CREATE AGGREGATE udf_avg(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + initcond = '{0,0}', + aggpartialfunc = udf_avg_p_int4_hasfinal +); +ERROR: finalfunc of aggpartialfunc must not be supplied when stype isn't internal +-- invalid: current user doesn't have execute privilege on aggpartialfunc +CREATE AGGREGATE udf_avg_p_int4(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + combinefunc = int4_avg_combine, + initcond = '{0,0}' +); +CREATE USER regress_priv_user1; +GRANT ALL ON SCHEMA public TO regress_priv_user1; +REVOKE EXECUTE ON FUNCTION udf_avg_p_int4 FROM public; +SET SESSION AUTHORIZATION regress_priv_user1; +CREATE AGGREGATE udf_avg_noprivilege(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + initcond = '{0,0}', + aggpartialfunc = udf_avg_p_int4 +); +ERROR: permission denied for function udf_avg_p_int4 +SET SESSION AUTHORIZATION DEFAULT; +REVOKE ALL ON SCHEMA public FROM regress_priv_user1; +DROP USER regress_priv_user1; +-- An aggregate function that is aggpartialfunc +-- of another aggregate cannot be deleted. +CREATE AGGREGATE udf_avg_int4_p(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + combinefunc = int4_avg_combine, + initcond = '{0,0}', + aggpartialfunc = udf_avg_int4_p +); +CREATE AGGREGATE udf_avg_int4(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + initcond = '{0,0}', + aggpartialfunc = udf_avg_int4_p +); +DROP AGGREGATE udf_avg_int4_p(int4); +ERROR: cannot drop function udf_avg_int4_p(integer) because other objects depend on it +DETAIL: function udf_avg_int4(integer) depends on function udf_avg_int4_p(integer) +HINT: Use DROP ... CASCADE to drop the dependent objects too. +DROP AGGREGATE udf_avg_int4_p(int4) CASCADE; +NOTICE: drop cascades to function udf_avg_int4(integer) diff --git a/src/test/regress/expected/oidjoins.out b/src/test/regress/expected/oidjoins.out index 215eb899be..6c80efa7eb 100644 --- a/src/test/regress/expected/oidjoins.out +++ b/src/test/regress/expected/oidjoins.out @@ -145,6 +145,7 @@ NOTICE: checking pg_aggregate {aggfinalfn} => pg_proc {oid} NOTICE: checking pg_aggregate {aggcombinefn} => pg_proc {oid} NOTICE: checking pg_aggregate {aggserialfn} => pg_proc {oid} NOTICE: checking pg_aggregate {aggdeserialfn} => pg_proc {oid} +NOTICE: checking pg_aggregate {aggpartialfn} => pg_proc {oid} NOTICE: checking pg_aggregate {aggmtransfn} => pg_proc {oid} NOTICE: checking pg_aggregate {aggminvtransfn} => pg_proc {oid} NOTICE: checking pg_aggregate {aggmfinalfn} => pg_proc {oid} diff --git a/src/test/regress/sql/create_aggregate.sql b/src/test/regress/sql/create_aggregate.sql index d4b4036fd7..e90e0ad2d6 100644 --- a/src/test/regress/sql/create_aggregate.sql +++ b/src/test/regress/sql/create_aggregate.sql @@ -328,3 +328,186 @@ CREATE AGGREGATE case_agg(float8) "Finalfunc_modify" = read_write, "Parallel" = safe ); + +-- invalid: finalfunc is specified and stype is not internal +-- even though aggpartialfunc is equal to the aggregate name +CREATE AGGREGATE udf_avg_int4_invalid(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + initcond = '{0,0}', + aggpartialfunc = udf_avg_int4_invalid +); + +-- invalid: finalfunc is not equal to serialfunc and stype is internal +-- even though aggpartialfunc is equal to the aggregate name +CREATE AGGREGATE udf_avg_int8_invalid(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + combinefunc = int8_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = udf_avg_int8_invalid +); + +-- invalid: aggpartialfunc which doesn't exist +CREATE AGGREGATE nonexistent_aggpartial_agg(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_sum, + combinefunc = int8_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = aggpartial_foo +); + +-- invalid: aggpartialfunc is specified and combinefunc isn't specified +CREATE AGGREGATE nocombinefunc_agg(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = avg_p_int8 +); + +-- invalid: combinefunc in agg and combinefunc in aggpartialfunc don't match +CREATE AGGREGATE udf_avg_invalid_combinefunc(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + combinefunc = numeric_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = avg_p_int8 +); + +CREATE FUNCTION udf_float8_accum(_float8, float8) RETURNS _float8 AS +$$ SELECT float8_accum($1, $2) $$ +LANGUAGE SQL; + +-- invalid: sfunc in agg and sfunc in aggpartialfunc don't match +CREATE AGGREGATE udf_avg_invalid_sfunc(float8) ( + sfunc = udf_float8_accum, + stype = _float8, + finalfunc = float8_avg, + combinefunc = float8_combine, + initcond = '{0,0,0}', + aggpartialfunc = avg_p_float8 +); +DROP FUNCTION udf_float8_accum(_float8, float8); + +-- invalid: initcond in agg and initcond in aggpartialfunc don't match +CREATE AGGREGATE udf_avg_invalid_initcond(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + initcond = '{1,0}', + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + aggpartialfunc = avg_p_int4 +); + +-- invalid: aggserialfn in agg and aggserialfn in aggpartialfunc don't match +CREATE AGGREGATE udf_avg_invalid_aggserialfn(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + combinefunc = int8_avg_combine, + serialfunc = numeric_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = avg_p_int8 +); + +-- invalid: aggdeserialfn in agg and aggdeserialfn in aggpartialfunc don't match +CREATE AGGREGATE udf_avg_invalid_aggdeserialfn(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + combinefunc = int8_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = numeric_avg_deserialize, + aggpartialfunc = avg_p_int8 +); + +-- invalid: stype is internal and aggpartialfunc's finalfunc +-- isn't serialfunc of agg +CREATE AGGREGATE udf_avg_p_int8(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_avg_serialize, + combinefunc = int8_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize +); +CREATE AGGREGATE udf_avg(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + combinefunc = int8_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = udf_avg_p_int8 +); + +-- invalid: stype is not internal and aggpartialfunc has finalfunc +CREATE AGGREGATE udf_avg_p_int4_hasfinal(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + initcond = '{0,0}' +); +CREATE AGGREGATE udf_avg(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + initcond = '{0,0}', + aggpartialfunc = udf_avg_p_int4_hasfinal +); + +-- invalid: current user doesn't have execute privilege on aggpartialfunc +CREATE AGGREGATE udf_avg_p_int4(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + combinefunc = int4_avg_combine, + initcond = '{0,0}' +); +CREATE USER regress_priv_user1; +GRANT ALL ON SCHEMA public TO regress_priv_user1; +REVOKE EXECUTE ON FUNCTION udf_avg_p_int4 FROM public; +SET SESSION AUTHORIZATION regress_priv_user1; + +CREATE AGGREGATE udf_avg_noprivilege(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + initcond = '{0,0}', + aggpartialfunc = udf_avg_p_int4 +); +SET SESSION AUTHORIZATION DEFAULT; +REVOKE ALL ON SCHEMA public FROM regress_priv_user1; +DROP USER regress_priv_user1; + +-- An aggregate function that is aggpartialfunc +-- of another aggregate cannot be deleted. +CREATE AGGREGATE udf_avg_int4_p(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + combinefunc = int4_avg_combine, + initcond = '{0,0}', + aggpartialfunc = udf_avg_int4_p +); +CREATE AGGREGATE udf_avg_int4(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + initcond = '{0,0}', + aggpartialfunc = udf_avg_int4_p +); +DROP AGGREGATE udf_avg_int4_p(int4); +DROP AGGREGATE udf_avg_int4_p(int4) CASCADE; -- 2.31.1 ^ permalink raw reply [nested|flat] 56+ messages in thread
* Re: Partial aggregates pushdown 2023-07-10 07:35 RE: Partial aggregates pushdown [email protected] <[email protected]> @ 2023-07-14 13:40 ` Alexander Pyhalov <[email protected]> 2023-07-18 01:35 ` RE: Partial aggregates pushdown [email protected] <[email protected]> 0 siblings, 1 reply; 56+ messages in thread From: Alexander Pyhalov @ 2023-07-14 13:40 UTC (permalink / raw) To: [email protected] <[email protected]>; +Cc: Bruce Momjian <[email protected]>; pgsql-hackers; Andres Freund <[email protected]>; Tom Lane <[email protected]>; Tomas Vondra <[email protected]>; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; Ilya Gladyshev <[email protected]> [email protected] писал 2023-07-10 10:35: > I have modified the program except for the point "if the version of > the remote server is less than PG17". > Instead, we have addressed the following. > "If check_partial_aggregate_support is true and the remote server > version is older than the local server > version, postgres_fdw does not assume that the partial aggregate > function is on the remote server unless > the partial aggregate function and the aggregate function match." > The reason for this is to maintain compatibility with any aggregate > function that does not support partial > aggregate in one version of V1 (V1 is PG17 or higher), even if the > next version supports partial aggregate. > For example, string_agg does not support partial aggregation in PG15, > but it will support partial aggregation > in PG16. > Hi. 1) In foreign_join_ok() should we set fpinfo->user if fpinfo->check_partial_aggregate_support is set like it's done for fpinfo->use_remote_estimate? It seems we can end up with fpinfo->user = NULL if use_remote_estimate is not set. 2) It seeems we found an additional issue with original patch, which is present in current one. I'm attaching a patch which seems to fix it, but I'm not quite sure in it. > We have not been able to add a test for the case where the remote > server version is older than the > local server version to the regression test. Is there any way to add > such tests to the existing regression > tests? > -- Best regards, Alexander Pyhalov, Postgres Professional Attachments: [text/x-diff] 0001-For-partial-aggregation-we-can-t-rely-on-the-fact-th.diff (11.8K, ../../[email protected]/2-0001-For-partial-aggregation-we-can-t-rely-on-the-fact-th.diff) download | inline diff: From 187d15185200aabc22c5219bbe636bc950670a78 Mon Sep 17 00:00:00 2001 From: Alexander Pyhalov <[email protected]> Date: Fri, 14 Jul 2023 16:34:02 +0300 Subject: [PATCH] For partial aggregation we can't rely on the fact that every var is a part of some GROUP BY expression --- .../postgres_fdw/expected/postgres_fdw.out | 120 ++++++++++++++++++ contrib/postgres_fdw/postgres_fdw.c | 31 +++-- contrib/postgres_fdw/sql/postgres_fdw.sql | 9 ++ 3 files changed, 152 insertions(+), 8 deletions(-) diff --git a/contrib/postgres_fdw/expected/postgres_fdw.out b/contrib/postgres_fdw/expected/postgres_fdw.out index 8d80ba0a6be..31ee461045c 100644 --- a/contrib/postgres_fdw/expected/postgres_fdw.out +++ b/contrib/postgres_fdw/expected/postgres_fdw.out @@ -9955,6 +9955,126 @@ SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; 49 | 19.0000000000000000 | 29 | 60 (50 rows) +EXPLAIN (VERBOSE, COSTS OFF) +SELECT b/2, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b/2 ORDER BY 1; + QUERY PLAN +-------------------------------------------------------------------------------------------------------------------------------- + Sort + Output: ((pagg_tab.b / 2)), (avg(pagg_tab.a)), (max(pagg_tab.a)), (count(*)) + Sort Key: ((pagg_tab.b / 2)) + -> Finalize HashAggregate + Output: ((pagg_tab.b / 2)), avg(pagg_tab.a), max(pagg_tab.a), count(*) + Group Key: ((pagg_tab.b / 2)) + -> Append + -> Foreign Scan + Output: ((pagg_tab.b / 2)), (PARTIAL avg(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab) + Remote SQL: SELECT (b / 2), avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p1 GROUP BY 1 + -> Foreign Scan + Output: ((pagg_tab_1.b / 2)), (PARTIAL avg(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1) + Remote SQL: SELECT (b / 2), avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p2 GROUP BY 1 + -> Foreign Scan + Output: ((pagg_tab_2.b / 2)), (PARTIAL avg(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2) + Remote SQL: SELECT (b / 2), avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p3 GROUP BY 1 +(19 rows) + +SELECT b/2, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b/2 ORDER BY 1; + ?column? | avg | max | count +----------+---------------------+-----+------- + 0 | 10.5000000000000000 | 21 | 120 + 1 | 12.5000000000000000 | 23 | 120 + 2 | 14.5000000000000000 | 25 | 120 + 3 | 16.5000000000000000 | 27 | 120 + 4 | 18.5000000000000000 | 29 | 120 + 5 | 10.5000000000000000 | 21 | 120 + 6 | 12.5000000000000000 | 23 | 120 + 7 | 14.5000000000000000 | 25 | 120 + 8 | 16.5000000000000000 | 27 | 120 + 9 | 18.5000000000000000 | 29 | 120 + 10 | 10.5000000000000000 | 21 | 120 + 11 | 12.5000000000000000 | 23 | 120 + 12 | 14.5000000000000000 | 25 | 120 + 13 | 16.5000000000000000 | 27 | 120 + 14 | 18.5000000000000000 | 29 | 120 + 15 | 10.5000000000000000 | 21 | 120 + 16 | 12.5000000000000000 | 23 | 120 + 17 | 14.5000000000000000 | 25 | 120 + 18 | 16.5000000000000000 | 27 | 120 + 19 | 18.5000000000000000 | 29 | 120 + 20 | 10.5000000000000000 | 21 | 120 + 21 | 12.5000000000000000 | 23 | 120 + 22 | 14.5000000000000000 | 25 | 120 + 23 | 16.5000000000000000 | 27 | 120 + 24 | 18.5000000000000000 | 29 | 120 +(25 rows) + +-- It's unsafe to pushdown variables if we need both variable and variable-based expression +EXPLAIN (VERBOSE, COSTS OFF) +SELECT (b/2)::numeric, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b/2 ORDER BY 1; + QUERY PLAN +---------------------------------------------------------------------------------------------------------------------------------------------- + Sort + Output: ((((pagg_tab.b / 2)))::numeric), (avg(pagg_tab.a)), (max(pagg_tab.a)), (count(*)), ((pagg_tab.b / 2)) + Sort Key: ((((pagg_tab.b / 2)))::numeric) + -> Finalize GroupAggregate + Output: (((pagg_tab.b / 2)))::numeric, avg(pagg_tab.a), max(pagg_tab.a), count(*), ((pagg_tab.b / 2)) + Group Key: ((pagg_tab.b / 2)) + -> Sort + Output: ((pagg_tab.b / 2)), pagg_tab.b, (PARTIAL avg(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)) + Sort Key: ((pagg_tab.b / 2)) + -> Append + -> Partial HashAggregate + Output: ((pagg_tab.b / 2)), pagg_tab.b, PARTIAL avg(pagg_tab.a), PARTIAL max(pagg_tab.a), PARTIAL count(*) + Group Key: (pagg_tab.b / 2) + -> Foreign Scan on public.fpagg_tab_p1 pagg_tab + Output: (pagg_tab.b / 2), pagg_tab.b, pagg_tab.a + Remote SQL: SELECT a, b FROM public.pagg_tab_p1 + -> Partial HashAggregate + Output: ((pagg_tab_1.b / 2)), pagg_tab_1.b, PARTIAL avg(pagg_tab_1.a), PARTIAL max(pagg_tab_1.a), PARTIAL count(*) + Group Key: (pagg_tab_1.b / 2) + -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1 + Output: (pagg_tab_1.b / 2), pagg_tab_1.b, pagg_tab_1.a + Remote SQL: SELECT a, b FROM public.pagg_tab_p2 + -> Partial HashAggregate + Output: ((pagg_tab_2.b / 2)), pagg_tab_2.b, PARTIAL avg(pagg_tab_2.a), PARTIAL max(pagg_tab_2.a), PARTIAL count(*) + Group Key: (pagg_tab_2.b / 2) + -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2 + Output: (pagg_tab_2.b / 2), pagg_tab_2.b, pagg_tab_2.a + Remote SQL: SELECT a, b FROM public.pagg_tab_p3 +(28 rows) + +SELECT (b/2)::numeric, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b/2 ORDER BY 1; + numeric | avg | max | count +---------+---------------------+-----+------- + 0 | 10.5000000000000000 | 21 | 120 + 1 | 12.5000000000000000 | 23 | 120 + 2 | 14.5000000000000000 | 25 | 120 + 3 | 16.5000000000000000 | 27 | 120 + 4 | 18.5000000000000000 | 29 | 120 + 5 | 10.5000000000000000 | 21 | 120 + 6 | 12.5000000000000000 | 23 | 120 + 7 | 14.5000000000000000 | 25 | 120 + 8 | 16.5000000000000000 | 27 | 120 + 9 | 18.5000000000000000 | 29 | 120 + 10 | 10.5000000000000000 | 21 | 120 + 11 | 12.5000000000000000 | 23 | 120 + 12 | 14.5000000000000000 | 25 | 120 + 13 | 16.5000000000000000 | 27 | 120 + 14 | 18.5000000000000000 | 29 | 120 + 15 | 10.5000000000000000 | 21 | 120 + 16 | 12.5000000000000000 | 23 | 120 + 17 | 14.5000000000000000 | 25 | 120 + 18 | 16.5000000000000000 | 27 | 120 + 19 | 18.5000000000000000 | 29 | 120 + 20 | 10.5000000000000000 | 21 | 120 + 21 | 12.5000000000000000 | 23 | 120 + 22 | 14.5000000000000000 | 25 | 120 + 23 | 16.5000000000000000 | 27 | 120 + 24 | 18.5000000000000000 | 29 | 120 +(25 rows) + -- Partial aggregates are safe to push down for all built-in aggregates EXPLAIN (VERBOSE, COSTS OFF) SELECT /* aggregate function <> aggpartialfunc */ diff --git a/contrib/postgres_fdw/postgres_fdw.c b/contrib/postgres_fdw/postgres_fdw.c index 5c48c5d50dc..eec76a00488 100644 --- a/contrib/postgres_fdw/postgres_fdw.c +++ b/contrib/postgres_fdw/postgres_fdw.c @@ -6381,13 +6381,17 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel, ListCell *lc; int i; List *tlist = NIL; + bool partial = false; /* We currently don't support pushing Grouping Sets. */ if (query->groupingSets) return false; + if (patype == PARTITIONWISE_AGGREGATE_PARTIAL) + partial = true; + /* It's unsafe to push having statements with partial aggregates */ - if ((patype == PARTITIONWISE_AGGREGATE_PARTIAL) && havingQual) + if (partial && havingQual) return false; /* Get the fpinfo of the underlying scan relation. */ @@ -6461,11 +6465,17 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel, } else { + /* + * For partial aggregation we can't rely on the fact that every + * var is a part of some GROUP BY expression, so extract variables + * and if we found some, avoid to push it down. + */ + /* * Non-grouping expression we need to compute. Can we ship it * as-is to the foreign server? */ - if (is_foreign_expr(root, grouped_rel, expr) && + if (!partial && is_foreign_expr(root, grouped_rel, expr) && !is_foreign_param(root, grouped_rel, expr)) { /* Yes, so add to tlist as-is; OK to suppress duplicates */ @@ -6489,14 +6499,17 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel, return false; /* - * Add aggregates, if any, into the targetlist. Plain Vars - * outside an aggregate can be ignored, because they should be + * Add aggregates, if any, into the targetlist. When dealing + * with usual aggregations, plain Vars outside an aggregate + * can be ignored, because they should be * either same as some GROUP BY column or part of some GROUP * BY expression. In either case, they are already part of - * the targetlist and thus no need to add them again. In fact - * including plain Vars in the tlist when they do not match a - * GROUP BY column would cause the foreign server to complain - * that the shipped query is invalid. + * the targetlist and thus no need to add them again. When + * looking at partial aggregation, we can get additional vars + * needed for partial aggregation itself, which do not match + * group by column. In fact including plain Vars in the tlist + * when they do not match a GROUP BY column would cause the + * foreign server to complain that the shipped query is invalid. */ foreach(l, aggvars) { @@ -6504,6 +6517,8 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel, if (IsA(aggref, Aggref)) tlist = add_to_flat_tlist(tlist, list_make1(aggref)); + else if (partial) + return false; } } } diff --git a/contrib/postgres_fdw/sql/postgres_fdw.sql b/contrib/postgres_fdw/sql/postgres_fdw.sql index 827637b001a..4f93e1931de 100644 --- a/contrib/postgres_fdw/sql/postgres_fdw.sql +++ b/contrib/postgres_fdw/sql/postgres_fdw.sql @@ -3030,6 +3030,15 @@ EXPLAIN (VERBOSE, COSTS OFF) SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; +EXPLAIN (VERBOSE, COSTS OFF) +SELECT b/2, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b/2 ORDER BY 1; +SELECT b/2, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b/2 ORDER BY 1; + +-- It's unsafe to pushdown variables if we need both variable and variable-based expression +EXPLAIN (VERBOSE, COSTS OFF) +SELECT (b/2)::numeric, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b/2 ORDER BY 1; +SELECT (b/2)::numeric, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b/2 ORDER BY 1; + -- Partial aggregates are safe to push down for all built-in aggregates EXPLAIN (VERBOSE, COSTS OFF) SELECT /* aggregate function <> aggpartialfunc */ -- 2.34.1 ^ permalink raw reply [nested|flat] 56+ messages in thread
* RE: Partial aggregates pushdown 2023-07-10 07:35 RE: Partial aggregates pushdown [email protected] <[email protected]> 2023-07-14 13:40 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]> @ 2023-07-18 01:35 ` [email protected] <[email protected]> 2023-07-19 00:43 ` RE: Partial aggregates pushdown [email protected] <[email protected]> 0 siblings, 1 reply; 56+ messages in thread From: [email protected] @ 2023-07-18 01:35 UTC (permalink / raw) To: Alexander Pyhalov <[email protected]>; +Cc: Bruce Momjian <[email protected]>; pgsql-hackers; Andres Freund <[email protected]>; Tom Lane <[email protected]>; Tomas Vondra <[email protected]>; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; Ilya Gladyshev <[email protected]>; [email protected] <[email protected]> Hi Mr.Pyhalov. > From: Alexander Pyhalov <[email protected]> > Sent: Friday, July 14, 2023 10:40 PM > 1) In foreign_join_ok() should we set fpinfo->user if > fpinfo->check_partial_aggregate_support is set like it's done for > fpinfo->use_remote_estimate? It seems we can end up with fpinfo->user > fpinfo->= > NULL if use_remote_estimate is not set. You are right. I will modify this patch according to your advice. Thank you for advice. > 2) It seeems we found an additional issue with original patch, which > is present in current one. I'm attaching a patch which seems to fix > it, but I'm not quite sure in it. Thank you for pointing out the issue. If a query's group-by clause contains variable based expression(not variable) and the query's select clause contains another expression, the partial aggregate could be unsafe to push down. An example of such queries: SELECT (b/2)::numeric, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b/2 Your patch disables partial aggregate pushdown for such queries. I'll see if we can modify the patch to safely do a partial aggregate pushdown for such queries as well. Such a query expects the variable in the select clause expression to be included in the target of the grouped rel (let see make_partial_grouping_target), but the original groupby clause has no reference to this variable, this seems to be the direct cause(let see foreign_grouping_ok). I will examine whether a safe pushdown can be achieved by matching the groupby clause information referenced by foreign_grouping_ok with the grouped rel target information. Sincerely yours, Yuuki Fujii -- Yuuki Fujii Information Technology R&D Center Mitsubishi Electric Corporation ^ permalink raw reply [nested|flat] 56+ messages in thread
* RE: Partial aggregates pushdown 2023-07-10 07:35 RE: Partial aggregates pushdown [email protected] <[email protected]> 2023-07-14 13:40 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]> 2023-07-18 01:35 ` RE: Partial aggregates pushdown [email protected] <[email protected]> @ 2023-07-19 00:43 ` [email protected] <[email protected]> 2023-07-20 10:23 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]> 0 siblings, 1 reply; 56+ messages in thread From: [email protected] @ 2023-07-19 00:43 UTC (permalink / raw) To: Alexander Pyhalov <[email protected]>; +Cc: Bruce Momjian <[email protected]>; pgsql-hackers; Andres Freund <[email protected]>; Tom Lane <[email protected]>; Tomas Vondra <[email protected]>; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; Ilya Gladyshev <[email protected]>; [email protected] <[email protected]> Hi Mr.Pyhalov, hackers. I have made the following three modifications about this patch. 1) > <[email protected]> > Sent: Tuesday, July 18, 2023 10:36 AM > > From: Alexander Pyhalov <[email protected]> > > Sent: Friday, July 14, 2023 10:40 PM > > 1) In foreign_join_ok() should we set fpinfo->user if > > fpinfo->check_partial_aggregate_support is set like it's done for > > fpinfo->use_remote_estimate? It seems we can end up with fpinfo->user > > fpinfo->= > > NULL if use_remote_estimate is not set. > You are right. I will modify this patch according to your advice. > Thank you for advice. Done. 2) > <[email protected]> > Sent: Tuesday, July 18, 2023 10:36 AM > > 2) It seeems we found an additional issue with original patch, which > > is present in current one. I'm attaching a patch which seems to fix > > it, but I'm not quite sure in it. > Thank you for pointing out the issue. > If a query's group-by clause contains variable based expression(not variable) > and the query's select clause contains another expression, > the partial aggregate could be unsafe to push down. > > An example of such queries: > SELECT (b/2)::numeric, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b/2 > > Your patch disables partial aggregate pushdown for such queries. > I'll see if we can modify the patch to safely do a partial aggregate pushdown for such queries as well. > Such a query expects the variable in the select clause expression to be included in the target of the grouped rel > (let see make_partial_grouping_target), > but the original groupby clause has no reference to this variable, > this seems to be the direct cause(let see foreign_grouping_ok). > I will examine whether a safe pushdown can be achieved by matching the > groupby clause information referenced by foreign_grouping_ok with the grouped rel target information. I modified the patch to safely do a partial aggregate pushdown for such queries as well by matching the groupby clause information referenced by foreign_grouping_ok with the grouped rel target information. 3) I modified the patch to safely do a partial aggregate pushdown for queries which contain having clauses. Sincerely yours, Yuuki Fujii -- Yuuki Fujii Information Technology R&D Center Mitsubishi Electric Corporation Attachments: [application/octet-stream] 0001-Partial-aggregates-push-down-v24.patch (264.7K, ../../OS3PR01MB66607B671A5C29DED7A41DF39539A@OS3PR01MB6660.jpnprd01.prod.outlook.com/2-0001-Partial-aggregates-push-down-v24.patch) download | inline diff: From: Yuki Fujii <[email protected]> Date: Tue, 18 Jul 2023 21:40:27 +0900 Subject: [PATCH] Partial aggregates push down v24 --- contrib/postgres_fdw/deparse.c | 103 ++- .../postgres_fdw/expected/postgres_fdw.out | 869 +++++++++++++++++- contrib/postgres_fdw/option.c | 4 +- contrib/postgres_fdw/postgres_fdw.c | 56 +- contrib/postgres_fdw/postgres_fdw.h | 10 + contrib/postgres_fdw/sql/postgres_fdw.sql | 311 ++++++- doc/src/sgml/catalogs.sgml | 13 + doc/src/sgml/postgres-fdw.sgml | 130 ++- doc/src/sgml/ref/create_aggregate.sgml | 42 + doc/src/sgml/xaggr.sgml | 15 +- src/backend/catalog/pg_aggregate.c | 101 ++ src/backend/commands/aggregatecmds.c | 4 + src/backend/optimizer/plan/planner.c | 69 +- src/bin/pg_dump/pg_dump.c | 9 +- src/include/catalog/pg_aggregate.dat | 680 +++++++++++--- src/include/catalog/pg_aggregate.h | 4 + src/include/catalog/pg_proc.dat | 196 ++++ src/include/nodes/pathnodes.h | 2 + .../regress/expected/create_aggregate.out | 184 ++++ src/test/regress/expected/oidjoins.out | 1 + src/test/regress/sql/create_aggregate.sql | 183 ++++ 21 files changed, 2782 insertions(+), 204 deletions(-) diff --git a/contrib/postgres_fdw/deparse.c b/contrib/postgres_fdw/deparse.c index 09d6dd60dd..e957611879 100644 --- a/contrib/postgres_fdw/deparse.c +++ b/contrib/postgres_fdw/deparse.c @@ -202,7 +202,7 @@ static bool is_subquery_var(Var *node, RelOptInfo *foreignrel, int *relno, int *colno); static void get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel, int *relno, int *colno); - +static bool partial_agg_ok(Aggref *agg, PgFdwRelationInfo *fpinfo); /* * Examine each qual clause in input_conds, and classify them into two groups, @@ -907,8 +907,9 @@ foreign_expr_walker(Node *node, if (!IS_UPPER_REL(glob_cxt->foreignrel)) return false; - /* Only non-split aggregates are pushable. */ - if (agg->aggsplit != AGGSPLIT_SIMPLE) + if ((agg->aggsplit != AGGSPLIT_SIMPLE) && (agg->aggsplit != AGGSPLIT_INITIAL_SERIAL)) + return false; + if (agg->aggsplit == AGGSPLIT_INITIAL_SERIAL && !partial_agg_ok(agg, fpinfo)) return false; /* As usual, it must be shippable. */ @@ -3517,14 +3518,34 @@ deparseAggref(Aggref *node, deparse_expr_cxt *context) StringInfo buf = context->buf; bool use_variadic; - /* Only basic, non-split aggregation accepted. */ - Assert(node->aggsplit == AGGSPLIT_SIMPLE); + + Assert((node->aggsplit == AGGSPLIT_SIMPLE) || + (node->aggsplit == AGGSPLIT_INITIAL_SERIAL)); /* Check if need to print VARIADIC (cf. ruleutils.c) */ use_variadic = node->aggvariadic; - /* Find aggregate name from aggfnoid which is a pg_proc entry */ - appendFunctionName(node->aggfnoid, context); + if (node->aggsplit == AGGSPLIT_SIMPLE) + /* Find aggregate name from aggfnoid which is a pg_proc entry */ + appendFunctionName(node->aggfnoid, context); + else + { + HeapTuple aggtup; + Form_pg_aggregate aggform; + + /* Find aggregate name from aggfnoid and aggpartialfn */ + aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(node->aggfnoid)); + if (!HeapTupleIsValid(aggtup)) + elog(ERROR, "cache lookup failed for aggregate %u", node->aggfnoid); + aggform = (Form_pg_aggregate) GETSTRUCT(aggtup); + + if (aggform->aggpartialfn) + appendFunctionName(aggform->aggpartialfn, context); + else + elog(ERROR, "there is no aggpartialfn %u", node->aggfnoid); + ReleaseSysCache(aggtup); + } + appendStringInfoChar(buf, '('); /* Add DISTINCT */ @@ -3723,6 +3744,7 @@ appendGroupByClause(List *tlist, deparse_expr_cxt *context) Query *query = context->root->parse; ListCell *lc; bool first = true; + PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) context->foreignrel->fdw_private; /* Nothing to be done, if there's no GROUP BY clause in the query. */ if (!query->groupClause) @@ -3743,7 +3765,7 @@ appendGroupByClause(List *tlist, deparse_expr_cxt *context) * to empty, and in any case the redundancy situation on the remote might * be different than what we think here. */ - foreach(lc, query->groupClause) + foreach(lc, fpinfo->group_clause) { SortGroupClause *grp = (SortGroupClause *) lfirst(lc); @@ -4047,3 +4069,68 @@ get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel, /* Shouldn't get here */ elog(ERROR, "unexpected expression in subquery output"); } + +/* + * Check that a buit-in aggpartialfunc exists on the remote server. If + * check_partial_aggregate_support is false, we assume the partial aggregate + * function exsits on the remote server. Otherwise we assume the partial + * aggregate function exsits on the remote server only if the remote server + * version is not less than the local server version. + */ +static bool +is_builtin_aggpartialfunc_shippable(Oid aggpartialfn, PgFdwRelationInfo *fpinfo) +{ + bool shippable = true; + + if (fpinfo->check_partial_aggregate_support) + { + if (fpinfo->remoteversion == 0) + { + PGconn *conn = GetConnection(fpinfo->user, false, NULL); + + fpinfo->remoteversion = PQserverVersion(conn); + } + if (fpinfo->remoteversion < PG_VERSION_NUM) + shippable = false; + } + return shippable; +} + +/* + * Check that partial aggregate agg is safe to push down. + * + * It is pushdown-safe when all of the following conditions are true. + * condition1) agg is AGGKIND_NORMAL aggregate which contains no distinct or + * order by clauses condition2) there is an aggregate function for partial + * aggregation (call it aggpartialfunc) corresponding to agg and + * aggpartialfunc exists on the remote server + */ +static bool +partial_agg_ok(Aggref *agg, PgFdwRelationInfo *fpinfo) +{ + HeapTuple aggtup; + Form_pg_aggregate aggform; + bool partial_agg_ok = true; + + Assert(agg->aggsplit == AGGSPLIT_INITIAL_SERIAL); + + /* We don't support complex partial aggregates */ + if (agg->aggdistinct || agg->aggkind != AGGKIND_NORMAL || agg->aggorder != NIL) + return false; + + aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(agg->aggfnoid)); + if (!HeapTupleIsValid(aggtup)) + elog(ERROR, "cache lookup failed for aggregate %u", agg->aggfnoid); + aggform = (Form_pg_aggregate) GETSTRUCT(aggtup); + + if (!aggform->aggpartialfn) + partial_agg_ok = false; + else if (is_builtin(aggform->aggpartialfn)) + partial_agg_ok = is_builtin_aggpartialfunc_shippable( + aggform->aggpartialfn, fpinfo); + else if (!is_shippable(aggform->aggpartialfn, ProcedureRelationId, fpinfo)) + partial_agg_ok = false; + + ReleaseSysCache(aggtup); + return partial_agg_ok; +} diff --git a/contrib/postgres_fdw/expected/postgres_fdw.out b/contrib/postgres_fdw/expected/postgres_fdw.out index c8c4614b54..3a0fa79151 100644 --- a/contrib/postgres_fdw/expected/postgres_fdw.out +++ b/contrib/postgres_fdw/expected/postgres_fdw.out @@ -9700,18 +9700,35 @@ RESET enable_partitionwise_join; -- =================================================================== -- test partitionwise aggregates -- =================================================================== -CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a); +ALTER SERVER loopback OPTIONS (ADD fdw_tuple_cost '0.1'); +CREATE TYPE mood AS enum ('sad', 'ok', 'happy'); +ALTER EXTENSION postgres_fdw ADD TYPE mood; +CREATE TABLE pagg_tab (a int, b int, c text, c_serial int4, + c_int4array _int4, c_interval interval, + c_money money, c_1c text, c_1b bytea, + c_bit bit(2), c_1or3int2 int2, + c_1or3int4 int4, c_1or3int8 int8, + c_bool bool, c_enum mood, c_pg_lsn pg_lsn, + c_tid tid, c_int4range int4range, + c_int4multirange int4multirange, + c_time time, c_timetz timetz, + c_timestamp timestamp, c_timestamptz timestamptz, + c_xid8 xid8) + PARTITION BY RANGE(a); CREATE TABLE pagg_tab_p1 (LIKE pagg_tab); CREATE TABLE pagg_tab_p2 (LIKE pagg_tab); CREATE TABLE pagg_tab_p3 (LIKE pagg_tab); -INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10; -INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10; -INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20; +INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i, array[(i % 2), 0], ((i % 2) || ' seconds')::interval, (i % 2)::money, ((i - 1) % 10)::text, (((i - 1) % 10)::text)::bytea, case when (i % 2) = 0 then B'01' else B'11' end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then true else false end, (case when (i % 2) = 0 then 'sad' else 'happy' end)::mood, ('0/' || i)::pg_lsn, ('(0,' || i || ')')::tid, int4range(0, i), int4multirange(int4range(0, i), int4range(100, 100 + i)), '00:00:00'::time + (i || ' seconds')::interval, '00:00:00'::timetz + (i || ' seconds')::interval, '2000-01-01'::timestamp + (i || ' seconds')::interval, '2000-01-01'::timestamptz + (i || ' seconds')::interval, ((i % 10)::text)::xid8 FROM generate_series(1, 3000) i WHERE (i % 30) < 10; +INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i, array[(i % 2), 0], ((i % 2) || ' seconds')::interval, (i % 2)::money, ((i - 1) % 10)::text, (((i - 1) % 10)::text)::bytea, case when (i % 2) = 0 then B'01' else B'11' end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then true else false end, (case when (i % 2) = 0 then 'sad' else 'happy' end)::mood, ('0/' || i)::pg_lsn, ('(0,' || i || ')')::tid, int4range(0, i), int4multirange(int4range(0, i), int4range(100, 100 + i)), '00:00:00'::time + (i || ' seconds')::interval, '00:00:00'::timetz + (i || ' seconds')::interval, '2000-01-01'::timestamp + (i || ' seconds')::interval, '2000-01-01'::timestamptz + (i || ' seconds')::interval, ((i % 10)::text)::xid8 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10; +INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i, array[(i % 2), 0], ((i % 2) || ' seconds')::interval, (i % 2)::money, ((i - 1) % 10)::text, (((i - 1) % 10)::text)::bytea, case when (i % 2) = 0 then B'01' else B'11' end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then true else false end, (case when (i % 2) = 0 then 'sad' else 'happy' end)::mood, ('0/' || i)::pg_lsn, ('(0,' || i || ')')::tid, int4range(0, i), int4multirange(int4range(0, i), int4range(100, 100 + i)), '00:00:00'::time + (i || ' seconds')::interval, '00:00:00'::timetz + (i || ' seconds')::interval, '2000-01-01'::timestamp + (i || ' seconds')::interval, '2000-01-01'::timestamptz + (i || ' seconds')::interval, ((i % 10)::text)::xid8 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20; -- Create foreign partitions CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1'); CREATE FOREIGN TABLE fpagg_tab_p2 PARTITION OF pagg_tab FOR VALUES FROM (10) TO (20) SERVER loopback OPTIONS (table_name 'pagg_tab_p2'); CREATE FOREIGN TABLE fpagg_tab_p3 PARTITION OF pagg_tab FOR VALUES FROM (20) TO (30) SERVER loopback OPTIONS (table_name 'pagg_tab_p3'); ANALYZE pagg_tab; +ANALYZE pagg_tab_p1; +ANALYZE pagg_tab_p2; +ANALYZE pagg_tab_p3; ANALYZE fpagg_tab_p1; ANALYZE fpagg_tab_p2; ANALYZE fpagg_tab_p3; @@ -9765,8 +9782,8 @@ SELECT a, sum(b), min(b), count(*) FROM pagg_tab GROUP BY a HAVING avg(b) < 22 O -- Should have all the columns in the target list for the given relation EXPLAIN (VERBOSE, COSTS OFF) SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1; - QUERY PLAN ------------------------------------------------------------------------- + QUERY PLAN +------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Sort Output: t1.a, (count(((t1.*)::pagg_tab))) Sort Key: t1.a @@ -9777,21 +9794,21 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1; Filter: (avg(t1.b) < '22'::numeric) -> Foreign Scan on public.fpagg_tab_p1 t1 Output: t1.a, t1.*, t1.b - Remote SQL: SELECT a, b, c FROM public.pagg_tab_p1 + Remote SQL: SELECT a, b, c, c_serial, c_int4array, c_interval, c_money, c_1c, c_1b, c_bit, c_1or3int2, c_1or3int4, c_1or3int8, c_bool, c_enum, c_pg_lsn, c_tid, c_int4range, c_int4multirange, c_time, c_timetz, c_timestamp, c_timestamptz, c_xid8 FROM public.pagg_tab_p1 -> HashAggregate Output: t1_1.a, count(((t1_1.*)::pagg_tab)) Group Key: t1_1.a Filter: (avg(t1_1.b) < '22'::numeric) -> Foreign Scan on public.fpagg_tab_p2 t1_1 Output: t1_1.a, t1_1.*, t1_1.b - Remote SQL: SELECT a, b, c FROM public.pagg_tab_p2 + Remote SQL: SELECT a, b, c, c_serial, c_int4array, c_interval, c_money, c_1c, c_1b, c_bit, c_1or3int2, c_1or3int4, c_1or3int8, c_bool, c_enum, c_pg_lsn, c_tid, c_int4range, c_int4multirange, c_time, c_timetz, c_timestamp, c_timestamptz, c_xid8 FROM public.pagg_tab_p2 -> HashAggregate Output: t1_2.a, count(((t1_2.*)::pagg_tab)) Group Key: t1_2.a Filter: (avg(t1_2.b) < '22'::numeric) -> Foreign Scan on public.fpagg_tab_p3 t1_2 Output: t1_2.a, t1_2.*, t1_2.b - Remote SQL: SELECT a, b, c FROM public.pagg_tab_p3 + Remote SQL: SELECT a, b, c, c_serial, c_int4array, c_interval, c_money, c_1c, c_1b, c_bit, c_1or3int2, c_1or3int4, c_1or3int8, c_bool, c_enum, c_pg_lsn, c_tid, c_int4range, c_int4multirange, c_time, c_timetz, c_timestamp, c_timestamptz, c_xid8 FROM public.pagg_tab_p3 (25 rows) SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1; @@ -9805,28 +9822,834 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1; 21 | 100 (6 rows) --- When GROUP BY clause does not match with PARTITION KEY. -EXPLAIN (COSTS OFF) +-- Partial aggregates are safe to push down having clause when there are partial aggregates +EXPLAIN (VERBOSE, COSTS OFF) SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1; - QUERY PLAN ------------------------------------------------------------------ + QUERY PLAN +----------------------------------------------------------------------------------------------------------------------------------------------------- + Finalize GroupAggregate + Output: pagg_tab.b, avg(pagg_tab.a), max(pagg_tab.a), count(*) + Group Key: pagg_tab.b + Filter: (sum(pagg_tab.a) < 700) + -> Sort + Output: pagg_tab.b, (PARTIAL avg(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.a)) + Sort Key: pagg_tab.b + -> Append + -> Foreign Scan + Output: pagg_tab.b, (PARTIAL avg(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.a)) + Filter: ((PARTIAL sum(pagg_tab.a)) < 700) + Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab) + Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*), sum(a) FROM public.pagg_tab_p1 GROUP BY 1 + -> Foreign Scan + Output: pagg_tab_1.b, (PARTIAL avg(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_1.a)) + Filter: ((PARTIAL sum(pagg_tab_1.a)) < 700) + Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1) + Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*), sum(a) FROM public.pagg_tab_p2 GROUP BY 1 + -> Foreign Scan + Output: pagg_tab_2.b, (PARTIAL avg(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_2.a)) + Filter: ((PARTIAL sum(pagg_tab_2.a)) < 700) + Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2) + Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*), sum(a) FROM public.pagg_tab_p3 GROUP BY 1 +(23 rows) + +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1; + b | avg | max | count +----+---------------------+-----+------- + 0 | 10.0000000000000000 | 20 | 60 + 1 | 11.0000000000000000 | 21 | 60 + 10 | 10.0000000000000000 | 20 | 60 + 11 | 11.0000000000000000 | 21 | 60 + 20 | 10.0000000000000000 | 20 | 60 + 21 | 11.0000000000000000 | 21 | 60 + 30 | 10.0000000000000000 | 20 | 60 + 31 | 11.0000000000000000 | 21 | 60 + 40 | 10.0000000000000000 | 20 | 60 + 41 | 11.0000000000000000 | 21 | 60 +(10 rows) + +-- Partial aggregates are safe to push down without having clause +EXPLAIN (VERBOSE, COSTS OFF) +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; + QUERY PLAN +------------------------------------------------------------------------------------------------------------------------ Sort + Output: pagg_tab.b, (avg(pagg_tab.a)), (max(pagg_tab.a)), (count(*)) Sort Key: pagg_tab.b -> Finalize HashAggregate + Output: pagg_tab.b, avg(pagg_tab.a), max(pagg_tab.a), count(*) Group Key: pagg_tab.b - Filter: (sum(pagg_tab.a) < 700) -> Append - -> Partial HashAggregate - Group Key: pagg_tab.b - -> Foreign Scan on fpagg_tab_p1 pagg_tab - -> Partial HashAggregate - Group Key: pagg_tab_1.b - -> Foreign Scan on fpagg_tab_p2 pagg_tab_1 - -> Partial HashAggregate - Group Key: pagg_tab_2.b - -> Foreign Scan on fpagg_tab_p3 pagg_tab_2 + -> Foreign Scan + Output: pagg_tab.b, (PARTIAL avg(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab) + Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p1 GROUP BY 1 + -> Foreign Scan + Output: pagg_tab_1.b, (PARTIAL avg(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1) + Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p2 GROUP BY 1 + -> Foreign Scan + Output: pagg_tab_2.b, (PARTIAL avg(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2) + Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p3 GROUP BY 1 +(19 rows) + +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; + b | avg | max | count +----+---------------------+-----+------- + 0 | 10.0000000000000000 | 20 | 60 + 1 | 11.0000000000000000 | 21 | 60 + 2 | 12.0000000000000000 | 22 | 60 + 3 | 13.0000000000000000 | 23 | 60 + 4 | 14.0000000000000000 | 24 | 60 + 5 | 15.0000000000000000 | 25 | 60 + 6 | 16.0000000000000000 | 26 | 60 + 7 | 17.0000000000000000 | 27 | 60 + 8 | 18.0000000000000000 | 28 | 60 + 9 | 19.0000000000000000 | 29 | 60 + 10 | 10.0000000000000000 | 20 | 60 + 11 | 11.0000000000000000 | 21 | 60 + 12 | 12.0000000000000000 | 22 | 60 + 13 | 13.0000000000000000 | 23 | 60 + 14 | 14.0000000000000000 | 24 | 60 + 15 | 15.0000000000000000 | 25 | 60 + 16 | 16.0000000000000000 | 26 | 60 + 17 | 17.0000000000000000 | 27 | 60 + 18 | 18.0000000000000000 | 28 | 60 + 19 | 19.0000000000000000 | 29 | 60 + 20 | 10.0000000000000000 | 20 | 60 + 21 | 11.0000000000000000 | 21 | 60 + 22 | 12.0000000000000000 | 22 | 60 + 23 | 13.0000000000000000 | 23 | 60 + 24 | 14.0000000000000000 | 24 | 60 + 25 | 15.0000000000000000 | 25 | 60 + 26 | 16.0000000000000000 | 26 | 60 + 27 | 17.0000000000000000 | 27 | 60 + 28 | 18.0000000000000000 | 28 | 60 + 29 | 19.0000000000000000 | 29 | 60 + 30 | 10.0000000000000000 | 20 | 60 + 31 | 11.0000000000000000 | 21 | 60 + 32 | 12.0000000000000000 | 22 | 60 + 33 | 13.0000000000000000 | 23 | 60 + 34 | 14.0000000000000000 | 24 | 60 + 35 | 15.0000000000000000 | 25 | 60 + 36 | 16.0000000000000000 | 26 | 60 + 37 | 17.0000000000000000 | 27 | 60 + 38 | 18.0000000000000000 | 28 | 60 + 39 | 19.0000000000000000 | 29 | 60 + 40 | 10.0000000000000000 | 20 | 60 + 41 | 11.0000000000000000 | 21 | 60 + 42 | 12.0000000000000000 | 22 | 60 + 43 | 13.0000000000000000 | 23 | 60 + 44 | 14.0000000000000000 | 24 | 60 + 45 | 15.0000000000000000 | 25 | 60 + 46 | 16.0000000000000000 | 26 | 60 + 47 | 17.0000000000000000 | 27 | 60 + 48 | 18.0000000000000000 | 28 | 60 + 49 | 19.0000000000000000 | 29 | 60 +(50 rows) + +-- Partial aggregates are safe to push down even if we need both variable and variable-based expression +EXPLAIN (VERBOSE, COSTS OFF) +SELECT (b/2)::numeric, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b/2 ORDER BY 1; + QUERY PLAN +---------------------------------------------------------------------------------------------------------------------------------------------- + Sort + Output: ((((pagg_tab.b / 2)))::numeric), (avg(pagg_tab.a)), (max(pagg_tab.a)), (count(*)), ((pagg_tab.b / 2)) + Sort Key: ((((pagg_tab.b / 2)))::numeric) + -> Finalize HashAggregate + Output: (((pagg_tab.b / 2)))::numeric, avg(pagg_tab.a), max(pagg_tab.a), count(*), ((pagg_tab.b / 2)) + Group Key: ((pagg_tab.b / 2)) + -> Append + -> Foreign Scan + Output: ((pagg_tab.b / 2)), pagg_tab.b, (PARTIAL avg(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab) + Remote SQL: SELECT (b / 2), b, avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p1 GROUP BY 1, 2 + -> Foreign Scan + Output: ((pagg_tab_1.b / 2)), pagg_tab_1.b, (PARTIAL avg(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1) + Remote SQL: SELECT (b / 2), b, avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p2 GROUP BY 1, 2 + -> Foreign Scan + Output: ((pagg_tab_2.b / 2)), pagg_tab_2.b, (PARTIAL avg(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2) + Remote SQL: SELECT (b / 2), b, avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p3 GROUP BY 1, 2 +(19 rows) + +SELECT (b/2)::numeric, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b/2 ORDER BY 1; + numeric | avg | max | count +---------+---------------------+-----+------- + 0 | 10.5000000000000000 | 21 | 120 + 1 | 12.5000000000000000 | 23 | 120 + 2 | 14.5000000000000000 | 25 | 120 + 3 | 16.5000000000000000 | 27 | 120 + 4 | 18.5000000000000000 | 29 | 120 + 5 | 10.5000000000000000 | 21 | 120 + 6 | 12.5000000000000000 | 23 | 120 + 7 | 14.5000000000000000 | 25 | 120 + 8 | 16.5000000000000000 | 27 | 120 + 9 | 18.5000000000000000 | 29 | 120 + 10 | 10.5000000000000000 | 21 | 120 + 11 | 12.5000000000000000 | 23 | 120 + 12 | 14.5000000000000000 | 25 | 120 + 13 | 16.5000000000000000 | 27 | 120 + 14 | 18.5000000000000000 | 29 | 120 + 15 | 10.5000000000000000 | 21 | 120 + 16 | 12.5000000000000000 | 23 | 120 + 17 | 14.5000000000000000 | 25 | 120 + 18 | 16.5000000000000000 | 27 | 120 + 19 | 18.5000000000000000 | 29 | 120 + 20 | 10.5000000000000000 | 21 | 120 + 21 | 12.5000000000000000 | 23 | 120 + 22 | 14.5000000000000000 | 25 | 120 + 23 | 16.5000000000000000 | 27 | 120 + 24 | 18.5000000000000000 | 29 | 120 +(25 rows) + +-- Partial aggregates are safe to push down for all built-in aggregates +EXPLAIN (VERBOSE, COSTS OFF) +SELECT /* aggregate function <> aggpartialfunc */ + array_agg(c_int4array), array_agg(b), + avg(b::int2), avg(b::int4), avg(b::int8), avg(c_interval), avg(b::float4), avg(b::float8), avg(b::numeric), + corr(b::float8, (b * b)::float8), + covar_pop(b::float8, (b * b)::float8), + covar_samp(b::float8, (b * b)::float8), + regr_avgx((2 * b)::float8, b::float8), + regr_avgy((2 * b)::float8, b::float8), + regr_intercept((2 * b + 3)::float8, b::float8), + regr_r2((b * b)::float8, b::float8), + regr_slope((2 * b + 3)::float8, b::float8), + regr_sxx((2 * b + 3)::float8, b::float8), + regr_sxy((b * b)::float8, b::float8), + regr_syy((2 * b + 3)::float8, b::float8), + stddev(b::float4), stddev(b::float8), stddev(b::int2), stddev(b::int4), stddev(b::int8), stddev(b::numeric), + stddev_pop(b::float4), stddev_pop(b::float8), stddev_pop(b::int2), stddev_pop(b::int4), stddev_pop(b::int8), stddev_pop(b::numeric), + stddev_samp(b::float4), stddev_samp(b::float8), stddev_samp(b::int2), stddev_samp(b::int4), stddev_samp(b::int8), stddev_samp(b::numeric), + string_agg(c_1c, ','), string_agg(c_1b, ','), + sum(b::int8), sum(b::numeric), + variance(b::float4), variance(b::float8), variance(b::int2), variance(b::int4), variance(b::int8), variance(b::numeric), + var_pop(b::float4), var_pop(b::float8), var_pop(b::int2), var_pop(b::int4), var_pop(b::int8), var_pop(b::numeric), + var_samp(b::float4), var_samp(b::float8), var_samp(b::int2), var_samp(b::int4), var_samp(b::int8), var_samp(b::numeric), + /* aggregate function = aggpartialfunc */ + any_value(b * 0), + bit_and(c_bit), bit_and(c_1or3int2), bit_and(c_1or3int4), bit_and(c_1or3int8), + bit_or(c_bit), bit_or(c_1or3int2), bit_or(c_1or3int4), bit_or(c_1or3int8), + bit_xor(c_bit), bit_xor(c_1or3int2), bit_xor(c_1or3int4), bit_xor(c_1or3int8), + bool_and(c_bool), + bool_or(c_bool), + count(b), count(*), + every(c_bool), + max(c_int4array), max(c_enum), max(c_1c::char(1)), max('2000-01-01'::date + b), max('0.0.0.0'::inet + b), max(b::float4), max(b::float8), max(b::int2), max(b::int4), max(b::int8), max(c_interval), max(c_money), max(b::numeric), max(b::oid), max(c_pg_lsn), max(c_tid), max(c_1c), max(c_time), max(c_timetz), max(c_timestamp), max(c_timestamptz), max(c_xid8), + min(c_int4array), min(c_enum), min(c_1c::char(1)), min('2000-01-01'::date + b), min('0.0.0.0'::inet + b), min(b::float4), min(b::float8), min(b::int2), min(b::int4), min(b::int8), min(c_interval), min(c_money), min(b::numeric), min(b::oid), min(c_pg_lsn), min(c_tid), min(c_1c), min(c_time), min(c_timetz), min(c_timestamp), min(c_timestamptz), min(c_xid8), + range_intersect_agg(c_int4range), range_intersect_agg(c_int4multirange), + regr_count((2 * b + 3)::float8, b::float8), + sum(b::float4), sum(b::float8), sum(b::int2), sum(b::int4), sum(c_interval), sum(c_money) + FROM pagg_tab WHERE c_serial between 1 and 30; + QUERY PLAN +------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- + Finalize Aggregate + Output: array_agg(pagg_tab.c_int4array), array_agg(pagg_tab.b), avg((pagg_tab.b)::smallint), avg(pagg_tab.b), avg((pagg_tab.b)::bigint), avg(pagg_tab.c_interval), avg((pagg_tab.b)::real), avg((pagg_tab.b)::double precision), avg((pagg_tab.b)::numeric), corr((pagg_tab.b)::double precision, ((pagg_tab.b * pagg_tab.b))::double precision), covar_pop((pagg_tab.b)::double precision, ((pagg_tab.b * pagg_tab.b))::double precision), covar_samp((pagg_tab.b)::double precision, ((pagg_tab.b * pagg_tab.b))::double precision), regr_avgx(((2 * pagg_tab.b))::double precision, (pagg_tab.b)::double precision), regr_avgy(((2 * pagg_tab.b))::double precision, (pagg_tab.b)::double precision), regr_intercept((((2 * pagg_tab.b) + 3))::double precision, (pagg_tab.b)::double precision), regr_r2(((pagg_tab.b * pagg_tab.b))::double precision, (pagg_tab.b)::double precision), regr_slope((((2 * pagg_tab.b) + 3))::double precision, (pagg_tab.b)::double precision), regr_sxx((((2 * pagg_tab.b) + 3))::double precision, (pagg_tab.b)::double precision), regr_sxy(((pagg_tab.b * pagg_tab.b))::double precision, (pagg_tab.b)::double precision), regr_syy((((2 * pagg_tab.b) + 3))::double precision, (pagg_tab.b)::double precision), stddev((pagg_tab.b)::real), stddev((pagg_tab.b)::double precision), stddev((pagg_tab.b)::smallint), stddev(pagg_tab.b), stddev((pagg_tab.b)::bigint), stddev((pagg_tab.b)::numeric), stddev_pop((pagg_tab.b)::real), stddev_pop((pagg_tab.b)::double precision), stddev_pop((pagg_tab.b)::smallint), stddev_pop(pagg_tab.b), stddev_pop((pagg_tab.b)::bigint), stddev_pop((pagg_tab.b)::numeric), stddev_samp((pagg_tab.b)::real), stddev_samp((pagg_tab.b)::double precision), stddev_samp((pagg_tab.b)::smallint), stddev_samp(pagg_tab.b), stddev_samp((pagg_tab.b)::bigint), stddev_samp((pagg_tab.b)::numeric), string_agg(pagg_tab.c_1c, ','::text), string_agg(pagg_tab.c_1b, '\x2c'::bytea), sum((pagg_tab.b)::bigint), sum((pagg_tab.b)::numeric), variance((pagg_tab.b)::real), variance((pagg_tab.b)::double precision), variance((pagg_tab.b)::smallint), variance(pagg_tab.b), variance((pagg_tab.b)::bigint), variance((pagg_tab.b)::numeric), var_pop((pagg_tab.b)::real), var_pop((pagg_tab.b)::double precision), var_pop((pagg_tab.b)::smallint), var_pop(pagg_tab.b), var_pop((pagg_tab.b)::bigint), var_pop((pagg_tab.b)::numeric), var_samp((pagg_tab.b)::real), var_samp((pagg_tab.b)::double precision), var_samp((pagg_tab.b)::smallint), var_samp(pagg_tab.b), var_samp((pagg_tab.b)::bigint), var_samp((pagg_tab.b)::numeric), any_value((pagg_tab.b * 0)), bit_and(pagg_tab.c_bit), bit_and(pagg_tab.c_1or3int2), bit_and(pagg_tab.c_1or3int4), bit_and(pagg_tab.c_1or3int8), bit_or(pagg_tab.c_bit), bit_or(pagg_tab.c_1or3int2), bit_or(pagg_tab.c_1or3int4), bit_or(pagg_tab.c_1or3int8), bit_xor(pagg_tab.c_bit), bit_xor(pagg_tab.c_1or3int2), bit_xor(pagg_tab.c_1or3int4), bit_xor(pagg_tab.c_1or3int8), bool_and(pagg_tab.c_bool), bool_or(pagg_tab.c_bool), count(pagg_tab.b), count(*), every(pagg_tab.c_bool), max(pagg_tab.c_int4array), max(pagg_tab.c_enum), max((pagg_tab.c_1c)::character(1)), max(('01-01-2000'::date + pagg_tab.b)), max(('0.0.0.0'::inet + (pagg_tab.b)::bigint)), max((pagg_tab.b)::real), max((pagg_tab.b)::double precision), max((pagg_tab.b)::smallint), max(pagg_tab.b), max((pagg_tab.b)::bigint), max(pagg_tab.c_interval), max(pagg_tab.c_money), max((pagg_tab.b)::numeric), max((pagg_tab.b)::oid), max(pagg_tab.c_pg_lsn), max(pagg_tab.c_tid), max(pagg_tab.c_1c), max(pagg_tab.c_time), max(pagg_tab.c_timetz), max(pagg_tab.c_timestamp), max(pagg_tab.c_timestamptz), max(pagg_tab.c_xid8), min(pagg_tab.c_int4array), min(pagg_tab.c_enum), min((pagg_tab.c_1c)::character(1)), min(('01-01-2000'::date + pagg_tab.b)), min(('0.0.0.0'::inet + (pagg_tab.b)::bigint)), min((pagg_tab.b)::real), min((pagg_tab.b)::double precision), min((pagg_tab.b)::smallint), min(pagg_tab.b), min((pagg_tab.b)::bigint), min(pagg_tab.c_interval), min(pagg_tab.c_money), min((pagg_tab.b)::numeric), min((pagg_tab.b)::oid), min(pagg_tab.c_pg_lsn), min(pagg_tab.c_tid), min(pagg_tab.c_1c), min(pagg_tab.c_time), min(pagg_tab.c_timetz), min(pagg_tab.c_timestamp), min(pagg_tab.c_timestamptz), min(pagg_tab.c_xid8), range_intersect_agg(pagg_tab.c_int4range), range_intersect_agg(pagg_tab.c_int4multirange), regr_count((((2 * pagg_tab.b) + 3))::double precision, (pagg_tab.b)::double precision), sum((pagg_tab.b)::real), sum((pagg_tab.b)::double precision), sum((pagg_tab.b)::smallint), sum(pagg_tab.b), sum(pagg_tab.c_interval), sum(pagg_tab.c_money) + -> Append + -> Foreign Scan + Output: (PARTIAL array_agg(pagg_tab.c_int4array)), (PARTIAL array_agg(pagg_tab.b)), (PARTIAL avg((pagg_tab.b)::smallint)), (PARTIAL avg(pagg_tab.b)), (PARTIAL avg((pagg_tab.b)::bigint)), (PARTIAL avg(pagg_tab.c_interval)), (PARTIAL avg((pagg_tab.b)::real)), (PARTIAL avg((pagg_tab.b)::double precision)), (PARTIAL avg((pagg_tab.b)::numeric)), (PARTIAL corr((pagg_tab.b)::double precision, ((pagg_tab.b * pagg_tab.b))::double precision)), (PARTIAL covar_pop((pagg_tab.b)::double precision, ((pagg_tab.b * pagg_tab.b))::double precision)), (PARTIAL covar_samp((pagg_tab.b)::double precision, ((pagg_tab.b * pagg_tab.b))::double precision)), (PARTIAL regr_avgx(((2 * pagg_tab.b))::double precision, (pagg_tab.b)::double precision)), (PARTIAL regr_avgy(((2 * pagg_tab.b))::double precision, (pagg_tab.b)::double precision)), (PARTIAL regr_intercept((((2 * pagg_tab.b) + 3))::double precision, (pagg_tab.b)::double precision)), (PARTIAL regr_r2(((pagg_tab.b * pagg_tab.b))::double precision, (pagg_tab.b)::double precision)), (PARTIAL regr_slope((((2 * pagg_tab.b) + 3))::double precision, (pagg_tab.b)::double precision)), (PARTIAL regr_sxx((((2 * pagg_tab.b) + 3))::double precision, (pagg_tab.b)::double precision)), (PARTIAL regr_sxy(((pagg_tab.b * pagg_tab.b))::double precision, (pagg_tab.b)::double precision)), (PARTIAL regr_syy((((2 * pagg_tab.b) + 3))::double precision, (pagg_tab.b)::double precision)), (PARTIAL stddev((pagg_tab.b)::real)), (PARTIAL stddev((pagg_tab.b)::double precision)), (PARTIAL stddev((pagg_tab.b)::smallint)), (PARTIAL stddev(pagg_tab.b)), (PARTIAL stddev((pagg_tab.b)::bigint)), (PARTIAL stddev((pagg_tab.b)::numeric)), (PARTIAL stddev_pop((pagg_tab.b)::real)), (PARTIAL stddev_pop((pagg_tab.b)::double precision)), (PARTIAL stddev_pop((pagg_tab.b)::smallint)), (PARTIAL stddev_pop(pagg_tab.b)), (PARTIAL stddev_pop((pagg_tab.b)::bigint)), (PARTIAL stddev_pop((pagg_tab.b)::numeric)), (PARTIAL stddev_samp((pagg_tab.b)::real)), (PARTIAL stddev_samp((pagg_tab.b)::double precision)), (PARTIAL stddev_samp((pagg_tab.b)::smallint)), (PARTIAL stddev_samp(pagg_tab.b)), (PARTIAL stddev_samp((pagg_tab.b)::bigint)), (PARTIAL stddev_samp((pagg_tab.b)::numeric)), (PARTIAL string_agg(pagg_tab.c_1c, ','::text)), (PARTIAL string_agg(pagg_tab.c_1b, '\x2c'::bytea)), (PARTIAL sum((pagg_tab.b)::bigint)), (PARTIAL sum((pagg_tab.b)::numeric)), (PARTIAL variance((pagg_tab.b)::real)), (PARTIAL variance((pagg_tab.b)::double precision)), (PARTIAL variance((pagg_tab.b)::smallint)), (PARTIAL variance(pagg_tab.b)), (PARTIAL variance((pagg_tab.b)::bigint)), (PARTIAL variance((pagg_tab.b)::numeric)), (PARTIAL var_pop((pagg_tab.b)::real)), (PARTIAL var_pop((pagg_tab.b)::double precision)), (PARTIAL var_pop((pagg_tab.b)::smallint)), (PARTIAL var_pop(pagg_tab.b)), (PARTIAL var_pop((pagg_tab.b)::bigint)), (PARTIAL var_pop((pagg_tab.b)::numeric)), (PARTIAL var_samp((pagg_tab.b)::real)), (PARTIAL var_samp((pagg_tab.b)::double precision)), (PARTIAL var_samp((pagg_tab.b)::smallint)), (PARTIAL var_samp(pagg_tab.b)), (PARTIAL var_samp((pagg_tab.b)::bigint)), (PARTIAL var_samp((pagg_tab.b)::numeric)), (PARTIAL any_value((pagg_tab.b * 0))), (PARTIAL bit_and(pagg_tab.c_bit)), (PARTIAL bit_and(pagg_tab.c_1or3int2)), (PARTIAL bit_and(pagg_tab.c_1or3int4)), (PARTIAL bit_and(pagg_tab.c_1or3int8)), (PARTIAL bit_or(pagg_tab.c_bit)), (PARTIAL bit_or(pagg_tab.c_1or3int2)), (PARTIAL bit_or(pagg_tab.c_1or3int4)), (PARTIAL bit_or(pagg_tab.c_1or3int8)), (PARTIAL bit_xor(pagg_tab.c_bit)), (PARTIAL bit_xor(pagg_tab.c_1or3int2)), (PARTIAL bit_xor(pagg_tab.c_1or3int4)), (PARTIAL bit_xor(pagg_tab.c_1or3int8)), (PARTIAL bool_and(pagg_tab.c_bool)), (PARTIAL bool_or(pagg_tab.c_bool)), (PARTIAL count(pagg_tab.b)), (PARTIAL count(*)), (PARTIAL every(pagg_tab.c_bool)), (PARTIAL max(pagg_tab.c_int4array)), (PARTIAL max(pagg_tab.c_enum)), (PARTIAL max((pagg_tab.c_1c)::character(1))), (PARTIAL max(('01-01-2000'::date + pagg_tab.b))), (PARTIAL max(('0.0.0.0'::inet + (pagg_tab.b)::bigint))), (PARTIAL max((pagg_tab.b)::real)), (PARTIAL max((pagg_tab.b)::double precision)), (PARTIAL max((pagg_tab.b)::smallint)), (PARTIAL max(pagg_tab.b)), (PARTIAL max((pagg_tab.b)::bigint)), (PARTIAL max(pagg_tab.c_interval)), (PARTIAL max(pagg_tab.c_money)), (PARTIAL max((pagg_tab.b)::numeric)), (PARTIAL max((pagg_tab.b)::oid)), (PARTIAL max(pagg_tab.c_pg_lsn)), (PARTIAL max(pagg_tab.c_tid)), (PARTIAL max(pagg_tab.c_1c)), (PARTIAL max(pagg_tab.c_time)), (PARTIAL max(pagg_tab.c_timetz)), (PARTIAL max(pagg_tab.c_timestamp)), (PARTIAL max(pagg_tab.c_timestamptz)), (PARTIAL max(pagg_tab.c_xid8)), (PARTIAL min(pagg_tab.c_int4array)), (PARTIAL min(pagg_tab.c_enum)), (PARTIAL min((pagg_tab.c_1c)::character(1))), (PARTIAL min(('01-01-2000'::date + pagg_tab.b))), (PARTIAL min(('0.0.0.0'::inet + (pagg_tab.b)::bigint))), (PARTIAL min((pagg_tab.b)::real)), (PARTIAL min((pagg_tab.b)::double precision)), (PARTIAL min((pagg_tab.b)::smallint)), (PARTIAL min(pagg_tab.b)), (PARTIAL min((pagg_tab.b)::bigint)), (PARTIAL min(pagg_tab.c_interval)), (PARTIAL min(pagg_tab.c_money)), (PARTIAL min((pagg_tab.b)::numeric)), (PARTIAL min((pagg_tab.b)::oid)), (PARTIAL min(pagg_tab.c_pg_lsn)), (PARTIAL min(pagg_tab.c_tid)), (PARTIAL min(pagg_tab.c_1c)), (PARTIAL min(pagg_tab.c_time)), (PARTIAL min(pagg_tab.c_timetz)), (PARTIAL min(pagg_tab.c_timestamp)), (PARTIAL min(pagg_tab.c_timestamptz)), (PARTIAL min(pagg_tab.c_xid8)), (PARTIAL range_intersect_agg(pagg_tab.c_int4range)), (PARTIAL range_intersect_agg(pagg_tab.c_int4multirange)), (PARTIAL regr_count((((2 * pagg_tab.b) + 3))::double precision, (pagg_tab.b)::double precision)), (PARTIAL sum((pagg_tab.b)::real)), (PARTIAL sum((pagg_tab.b)::double precision)), (PARTIAL sum((pagg_tab.b)::smallint)), (PARTIAL sum(pagg_tab.b)), (PARTIAL sum(pagg_tab.c_interval)), (PARTIAL sum(pagg_tab.c_money)) + Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab) + Remote SQL: SELECT array_agg_p_anyarray(c_int4array), array_agg_p_anynonarray(b), avg_p_int2(b::smallint), avg_p_int4(b), avg_p_int8(b::bigint), avg_p_interval(c_interval), avg_p_float4(b::real), avg_p_float8(b::double precision), avg_p_numeric(b::numeric), corr_p(b::double precision, (b * b)::double precision), covar_pop_p(b::double precision, (b * b)::double precision), covar_samp_p(b::double precision, (b * b)::double precision), regr_avgx_p((2 * b)::double precision, b::double precision), regr_avgy_p((2 * b)::double precision, b::double precision), regr_intercept_p(((2 * b) + 3)::double precision, b::double precision), regr_r2_p((b * b)::double precision, b::double precision), regr_slope_p(((2 * b) + 3)::double precision, b::double precision), regr_sxx_p(((2 * b) + 3)::double precision, b::double precision), regr_sxy_p((b * b)::double precision, b::double precision), regr_syy_p(((2 * b) + 3)::double precision, b::double precision), stddev_p_float4(b::real), stddev_p_float8(b::double precision), stddev_p_int2(b::smallint), stddev_p_int4(b), stddev_p_int8(b::bigint), stddev_p_numeric(b::numeric), stddev_pop_p_float4(b::real), stddev_pop_p_float8(b::double precision), stddev_pop_p_int2(b::smallint), stddev_pop_p_int4(b), stddev_pop_p_int8(b::bigint), stddev_pop_p_numeric(b::numeric), stddev_samp_p_float4(b::real), stddev_samp_p_float8(b::double precision), stddev_samp_p_int2(b::smallint), stddev_samp_p_int4(b), stddev_samp_p_int8(b::bigint), stddev_samp_p_numeric(b::numeric), string_agg_p_text_text(c_1c, ','::text), string_agg_p_bytea_bytea(c_1b, E'\\x2c'::bytea), sum_p_int8(b::bigint), sum_p_numeric(b::numeric), variance_p_float4(b::real), variance_p_float8(b::double precision), variance_p_int2(b::smallint), variance_p_int4(b), variance_p_int8(b::bigint), variance_p_numeric(b::numeric), var_pop_p_float4(b::real), var_pop_p_float8(b::double precision), var_pop_p_int2(b::smallint), var_pop_p_int4(b), var_pop_p_int8(b::bigint), var_pop_p_numeric(b::numeric), var_samp_p_float4(b::real), var_samp_p_float8(b::double precision), var_samp_p_int2(b::smallint), var_samp_p_int4(b), var_samp_p_int8(b::bigint), var_samp_p_numeric(b::numeric), any_value((b * 0)), bit_and(c_bit), bit_and(c_1or3int2), bit_and(c_1or3int4), bit_and(c_1or3int8), bit_or(c_bit), bit_or(c_1or3int2), bit_or(c_1or3int4), bit_or(c_1or3int8), bit_xor(c_bit), bit_xor(c_1or3int2), bit_xor(c_1or3int4), bit_xor(c_1or3int8), bool_and(c_bool), bool_or(c_bool), count(b), count(*), every(c_bool), max(c_int4array), max(c_enum), max(c_1c::character(1)), max(('01-01-2000'::date + b)), max(('0.0.0.0'::inet + b)), max(b::real), max(b::double precision), max(b::smallint), max(b), max(b::bigint), max(c_interval), max(c_money), max(b::numeric), max(b::oid), max(c_pg_lsn), max(c_tid), max(c_1c), max(c_time), max(c_timetz), max(c_timestamp), max(c_timestamptz), max(c_xid8), min(c_int4array), min(c_enum), min(c_1c::character(1)), min(('01-01-2000'::date + b)), min(('0.0.0.0'::inet + b)), min(b::real), min(b::double precision), min(b::smallint), min(b), min(b::bigint), min(c_interval), min(c_money), min(b::numeric), min(b::oid), min(c_pg_lsn), min(c_tid), min(c_1c), min(c_time), min(c_timetz), min(c_timestamp), min(c_timestamptz), min(c_xid8), range_intersect_agg(c_int4range), range_intersect_agg(c_int4multirange), regr_count(((2 * b) + 3)::double precision, b::double precision), sum(b::real), sum(b::double precision), sum(b::smallint), sum(b), sum(c_interval), sum(c_money) FROM public.pagg_tab_p1 WHERE ((c_serial >= 1)) AND ((c_serial <= 30)) + -> Foreign Scan + Output: (PARTIAL array_agg(pagg_tab_1.c_int4array)), (PARTIAL array_agg(pagg_tab_1.b)), (PARTIAL avg((pagg_tab_1.b)::smallint)), (PARTIAL avg(pagg_tab_1.b)), (PARTIAL avg((pagg_tab_1.b)::bigint)), (PARTIAL avg(pagg_tab_1.c_interval)), (PARTIAL avg((pagg_tab_1.b)::real)), (PARTIAL avg((pagg_tab_1.b)::double precision)), (PARTIAL avg((pagg_tab_1.b)::numeric)), (PARTIAL corr((pagg_tab_1.b)::double precision, ((pagg_tab_1.b * pagg_tab_1.b))::double precision)), (PARTIAL covar_pop((pagg_tab_1.b)::double precision, ((pagg_tab_1.b * pagg_tab_1.b))::double precision)), (PARTIAL covar_samp((pagg_tab_1.b)::double precision, ((pagg_tab_1.b * pagg_tab_1.b))::double precision)), (PARTIAL regr_avgx(((2 * pagg_tab_1.b))::double precision, (pagg_tab_1.b)::double precision)), (PARTIAL regr_avgy(((2 * pagg_tab_1.b))::double precision, (pagg_tab_1.b)::double precision)), (PARTIAL regr_intercept((((2 * pagg_tab_1.b) + 3))::double precision, (pagg_tab_1.b)::double precision)), (PARTIAL regr_r2(((pagg_tab_1.b * pagg_tab_1.b))::double precision, (pagg_tab_1.b)::double precision)), (PARTIAL regr_slope((((2 * pagg_tab_1.b) + 3))::double precision, (pagg_tab_1.b)::double precision)), (PARTIAL regr_sxx((((2 * pagg_tab_1.b) + 3))::double precision, (pagg_tab_1.b)::double precision)), (PARTIAL regr_sxy(((pagg_tab_1.b * pagg_tab_1.b))::double precision, (pagg_tab_1.b)::double precision)), (PARTIAL regr_syy((((2 * pagg_tab_1.b) + 3))::double precision, (pagg_tab_1.b)::double precision)), (PARTIAL stddev((pagg_tab_1.b)::real)), (PARTIAL stddev((pagg_tab_1.b)::double precision)), (PARTIAL stddev((pagg_tab_1.b)::smallint)), (PARTIAL stddev(pagg_tab_1.b)), (PARTIAL stddev((pagg_tab_1.b)::bigint)), (PARTIAL stddev((pagg_tab_1.b)::numeric)), (PARTIAL stddev_pop((pagg_tab_1.b)::real)), (PARTIAL stddev_pop((pagg_tab_1.b)::double precision)), (PARTIAL stddev_pop((pagg_tab_1.b)::smallint)), (PARTIAL stddev_pop(pagg_tab_1.b)), (PARTIAL stddev_pop((pagg_tab_1.b)::bigint)), (PARTIAL stddev_pop((pagg_tab_1.b)::numeric)), (PARTIAL stddev_samp((pagg_tab_1.b)::real)), (PARTIAL stddev_samp((pagg_tab_1.b)::double precision)), (PARTIAL stddev_samp((pagg_tab_1.b)::smallint)), (PARTIAL stddev_samp(pagg_tab_1.b)), (PARTIAL stddev_samp((pagg_tab_1.b)::bigint)), (PARTIAL stddev_samp((pagg_tab_1.b)::numeric)), (PARTIAL string_agg(pagg_tab_1.c_1c, ','::text)), (PARTIAL string_agg(pagg_tab_1.c_1b, '\x2c'::bytea)), (PARTIAL sum((pagg_tab_1.b)::bigint)), (PARTIAL sum((pagg_tab_1.b)::numeric)), (PARTIAL variance((pagg_tab_1.b)::real)), (PARTIAL variance((pagg_tab_1.b)::double precision)), (PARTIAL variance((pagg_tab_1.b)::smallint)), (PARTIAL variance(pagg_tab_1.b)), (PARTIAL variance((pagg_tab_1.b)::bigint)), (PARTIAL variance((pagg_tab_1.b)::numeric)), (PARTIAL var_pop((pagg_tab_1.b)::real)), (PARTIAL var_pop((pagg_tab_1.b)::double precision)), (PARTIAL var_pop((pagg_tab_1.b)::smallint)), (PARTIAL var_pop(pagg_tab_1.b)), (PARTIAL var_pop((pagg_tab_1.b)::bigint)), (PARTIAL var_pop((pagg_tab_1.b)::numeric)), (PARTIAL var_samp((pagg_tab_1.b)::real)), (PARTIAL var_samp((pagg_tab_1.b)::double precision)), (PARTIAL var_samp((pagg_tab_1.b)::smallint)), (PARTIAL var_samp(pagg_tab_1.b)), (PARTIAL var_samp((pagg_tab_1.b)::bigint)), (PARTIAL var_samp((pagg_tab_1.b)::numeric)), (PARTIAL any_value((pagg_tab_1.b * 0))), (PARTIAL bit_and(pagg_tab_1.c_bit)), (PARTIAL bit_and(pagg_tab_1.c_1or3int2)), (PARTIAL bit_and(pagg_tab_1.c_1or3int4)), (PARTIAL bit_and(pagg_tab_1.c_1or3int8)), (PARTIAL bit_or(pagg_tab_1.c_bit)), (PARTIAL bit_or(pagg_tab_1.c_1or3int2)), (PARTIAL bit_or(pagg_tab_1.c_1or3int4)), (PARTIAL bit_or(pagg_tab_1.c_1or3int8)), (PARTIAL bit_xor(pagg_tab_1.c_bit)), (PARTIAL bit_xor(pagg_tab_1.c_1or3int2)), (PARTIAL bit_xor(pagg_tab_1.c_1or3int4)), (PARTIAL bit_xor(pagg_tab_1.c_1or3int8)), (PARTIAL bool_and(pagg_tab_1.c_bool)), (PARTIAL bool_or(pagg_tab_1.c_bool)), (PARTIAL count(pagg_tab_1.b)), (PARTIAL count(*)), (PARTIAL every(pagg_tab_1.c_bool)), (PARTIAL max(pagg_tab_1.c_int4array)), (PARTIAL max(pagg_tab_1.c_enum)), (PARTIAL max((pagg_tab_1.c_1c)::character(1))), (PARTIAL max(('01-01-2000'::date + pagg_tab_1.b))), (PARTIAL max(('0.0.0.0'::inet + (pagg_tab_1.b)::bigint))), (PARTIAL max((pagg_tab_1.b)::real)), (PARTIAL max((pagg_tab_1.b)::double precision)), (PARTIAL max((pagg_tab_1.b)::smallint)), (PARTIAL max(pagg_tab_1.b)), (PARTIAL max((pagg_tab_1.b)::bigint)), (PARTIAL max(pagg_tab_1.c_interval)), (PARTIAL max(pagg_tab_1.c_money)), (PARTIAL max((pagg_tab_1.b)::numeric)), (PARTIAL max((pagg_tab_1.b)::oid)), (PARTIAL max(pagg_tab_1.c_pg_lsn)), (PARTIAL max(pagg_tab_1.c_tid)), (PARTIAL max(pagg_tab_1.c_1c)), (PARTIAL max(pagg_tab_1.c_time)), (PARTIAL max(pagg_tab_1.c_timetz)), (PARTIAL max(pagg_tab_1.c_timestamp)), (PARTIAL max(pagg_tab_1.c_timestamptz)), (PARTIAL max(pagg_tab_1.c_xid8)), (PARTIAL min(pagg_tab_1.c_int4array)), (PARTIAL min(pagg_tab_1.c_enum)), (PARTIAL min((pagg_tab_1.c_1c)::character(1))), (PARTIAL min(('01-01-2000'::date + pagg_tab_1.b))), (PARTIAL min(('0.0.0.0'::inet + (pagg_tab_1.b)::bigint))), (PARTIAL min((pagg_tab_1.b)::real)), (PARTIAL min((pagg_tab_1.b)::double precision)), (PARTIAL min((pagg_tab_1.b)::smallint)), (PARTIAL min(pagg_tab_1.b)), (PARTIAL min((pagg_tab_1.b)::bigint)), (PARTIAL min(pagg_tab_1.c_interval)), (PARTIAL min(pagg_tab_1.c_money)), (PARTIAL min((pagg_tab_1.b)::numeric)), (PARTIAL min((pagg_tab_1.b)::oid)), (PARTIAL min(pagg_tab_1.c_pg_lsn)), (PARTIAL min(pagg_tab_1.c_tid)), (PARTIAL min(pagg_tab_1.c_1c)), (PARTIAL min(pagg_tab_1.c_time)), (PARTIAL min(pagg_tab_1.c_timetz)), (PARTIAL min(pagg_tab_1.c_timestamp)), (PARTIAL min(pagg_tab_1.c_timestamptz)), (PARTIAL min(pagg_tab_1.c_xid8)), (PARTIAL range_intersect_agg(pagg_tab_1.c_int4range)), (PARTIAL range_intersect_agg(pagg_tab_1.c_int4multirange)), (PARTIAL regr_count((((2 * pagg_tab_1.b) + 3))::double precision, (pagg_tab_1.b)::double precision)), (PARTIAL sum((pagg_tab_1.b)::real)), (PARTIAL sum((pagg_tab_1.b)::double precision)), (PARTIAL sum((pagg_tab_1.b)::smallint)), (PARTIAL sum(pagg_tab_1.b)), (PARTIAL sum(pagg_tab_1.c_interval)), (PARTIAL sum(pagg_tab_1.c_money)) + Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1) + Remote SQL: SELECT array_agg_p_anyarray(c_int4array), array_agg_p_anynonarray(b), avg_p_int2(b::smallint), avg_p_int4(b), avg_p_int8(b::bigint), avg_p_interval(c_interval), avg_p_float4(b::real), avg_p_float8(b::double precision), avg_p_numeric(b::numeric), corr_p(b::double precision, (b * b)::double precision), covar_pop_p(b::double precision, (b * b)::double precision), covar_samp_p(b::double precision, (b * b)::double precision), regr_avgx_p((2 * b)::double precision, b::double precision), regr_avgy_p((2 * b)::double precision, b::double precision), regr_intercept_p(((2 * b) + 3)::double precision, b::double precision), regr_r2_p((b * b)::double precision, b::double precision), regr_slope_p(((2 * b) + 3)::double precision, b::double precision), regr_sxx_p(((2 * b) + 3)::double precision, b::double precision), regr_sxy_p((b * b)::double precision, b::double precision), regr_syy_p(((2 * b) + 3)::double precision, b::double precision), stddev_p_float4(b::real), stddev_p_float8(b::double precision), stddev_p_int2(b::smallint), stddev_p_int4(b), stddev_p_int8(b::bigint), stddev_p_numeric(b::numeric), stddev_pop_p_float4(b::real), stddev_pop_p_float8(b::double precision), stddev_pop_p_int2(b::smallint), stddev_pop_p_int4(b), stddev_pop_p_int8(b::bigint), stddev_pop_p_numeric(b::numeric), stddev_samp_p_float4(b::real), stddev_samp_p_float8(b::double precision), stddev_samp_p_int2(b::smallint), stddev_samp_p_int4(b), stddev_samp_p_int8(b::bigint), stddev_samp_p_numeric(b::numeric), string_agg_p_text_text(c_1c, ','::text), string_agg_p_bytea_bytea(c_1b, E'\\x2c'::bytea), sum_p_int8(b::bigint), sum_p_numeric(b::numeric), variance_p_float4(b::real), variance_p_float8(b::double precision), variance_p_int2(b::smallint), variance_p_int4(b), variance_p_int8(b::bigint), variance_p_numeric(b::numeric), var_pop_p_float4(b::real), var_pop_p_float8(b::double precision), var_pop_p_int2(b::smallint), var_pop_p_int4(b), var_pop_p_int8(b::bigint), var_pop_p_numeric(b::numeric), var_samp_p_float4(b::real), var_samp_p_float8(b::double precision), var_samp_p_int2(b::smallint), var_samp_p_int4(b), var_samp_p_int8(b::bigint), var_samp_p_numeric(b::numeric), any_value((b * 0)), bit_and(c_bit), bit_and(c_1or3int2), bit_and(c_1or3int4), bit_and(c_1or3int8), bit_or(c_bit), bit_or(c_1or3int2), bit_or(c_1or3int4), bit_or(c_1or3int8), bit_xor(c_bit), bit_xor(c_1or3int2), bit_xor(c_1or3int4), bit_xor(c_1or3int8), bool_and(c_bool), bool_or(c_bool), count(b), count(*), every(c_bool), max(c_int4array), max(c_enum), max(c_1c::character(1)), max(('01-01-2000'::date + b)), max(('0.0.0.0'::inet + b)), max(b::real), max(b::double precision), max(b::smallint), max(b), max(b::bigint), max(c_interval), max(c_money), max(b::numeric), max(b::oid), max(c_pg_lsn), max(c_tid), max(c_1c), max(c_time), max(c_timetz), max(c_timestamp), max(c_timestamptz), max(c_xid8), min(c_int4array), min(c_enum), min(c_1c::character(1)), min(('01-01-2000'::date + b)), min(('0.0.0.0'::inet + b)), min(b::real), min(b::double precision), min(b::smallint), min(b), min(b::bigint), min(c_interval), min(c_money), min(b::numeric), min(b::oid), min(c_pg_lsn), min(c_tid), min(c_1c), min(c_time), min(c_timetz), min(c_timestamp), min(c_timestamptz), min(c_xid8), range_intersect_agg(c_int4range), range_intersect_agg(c_int4multirange), regr_count(((2 * b) + 3)::double precision, b::double precision), sum(b::real), sum(b::double precision), sum(b::smallint), sum(b), sum(c_interval), sum(c_money) FROM public.pagg_tab_p2 WHERE ((c_serial >= 1)) AND ((c_serial <= 30)) + -> Foreign Scan + Output: (PARTIAL array_agg(pagg_tab_2.c_int4array)), (PARTIAL array_agg(pagg_tab_2.b)), (PARTIAL avg((pagg_tab_2.b)::smallint)), (PARTIAL avg(pagg_tab_2.b)), (PARTIAL avg((pagg_tab_2.b)::bigint)), (PARTIAL avg(pagg_tab_2.c_interval)), (PARTIAL avg((pagg_tab_2.b)::real)), (PARTIAL avg((pagg_tab_2.b)::double precision)), (PARTIAL avg((pagg_tab_2.b)::numeric)), (PARTIAL corr((pagg_tab_2.b)::double precision, ((pagg_tab_2.b * pagg_tab_2.b))::double precision)), (PARTIAL covar_pop((pagg_tab_2.b)::double precision, ((pagg_tab_2.b * pagg_tab_2.b))::double precision)), (PARTIAL covar_samp((pagg_tab_2.b)::double precision, ((pagg_tab_2.b * pagg_tab_2.b))::double precision)), (PARTIAL regr_avgx(((2 * pagg_tab_2.b))::double precision, (pagg_tab_2.b)::double precision)), (PARTIAL regr_avgy(((2 * pagg_tab_2.b))::double precision, (pagg_tab_2.b)::double precision)), (PARTIAL regr_intercept((((2 * pagg_tab_2.b) + 3))::double precision, (pagg_tab_2.b)::double precision)), (PARTIAL regr_r2(((pagg_tab_2.b * pagg_tab_2.b))::double precision, (pagg_tab_2.b)::double precision)), (PARTIAL regr_slope((((2 * pagg_tab_2.b) + 3))::double precision, (pagg_tab_2.b)::double precision)), (PARTIAL regr_sxx((((2 * pagg_tab_2.b) + 3))::double precision, (pagg_tab_2.b)::double precision)), (PARTIAL regr_sxy(((pagg_tab_2.b * pagg_tab_2.b))::double precision, (pagg_tab_2.b)::double precision)), (PARTIAL regr_syy((((2 * pagg_tab_2.b) + 3))::double precision, (pagg_tab_2.b)::double precision)), (PARTIAL stddev((pagg_tab_2.b)::real)), (PARTIAL stddev((pagg_tab_2.b)::double precision)), (PARTIAL stddev((pagg_tab_2.b)::smallint)), (PARTIAL stddev(pagg_tab_2.b)), (PARTIAL stddev((pagg_tab_2.b)::bigint)), (PARTIAL stddev((pagg_tab_2.b)::numeric)), (PARTIAL stddev_pop((pagg_tab_2.b)::real)), (PARTIAL stddev_pop((pagg_tab_2.b)::double precision)), (PARTIAL stddev_pop((pagg_tab_2.b)::smallint)), (PARTIAL stddev_pop(pagg_tab_2.b)), (PARTIAL stddev_pop((pagg_tab_2.b)::bigint)), (PARTIAL stddev_pop((pagg_tab_2.b)::numeric)), (PARTIAL stddev_samp((pagg_tab_2.b)::real)), (PARTIAL stddev_samp((pagg_tab_2.b)::double precision)), (PARTIAL stddev_samp((pagg_tab_2.b)::smallint)), (PARTIAL stddev_samp(pagg_tab_2.b)), (PARTIAL stddev_samp((pagg_tab_2.b)::bigint)), (PARTIAL stddev_samp((pagg_tab_2.b)::numeric)), (PARTIAL string_agg(pagg_tab_2.c_1c, ','::text)), (PARTIAL string_agg(pagg_tab_2.c_1b, '\x2c'::bytea)), (PARTIAL sum((pagg_tab_2.b)::bigint)), (PARTIAL sum((pagg_tab_2.b)::numeric)), (PARTIAL variance((pagg_tab_2.b)::real)), (PARTIAL variance((pagg_tab_2.b)::double precision)), (PARTIAL variance((pagg_tab_2.b)::smallint)), (PARTIAL variance(pagg_tab_2.b)), (PARTIAL variance((pagg_tab_2.b)::bigint)), (PARTIAL variance((pagg_tab_2.b)::numeric)), (PARTIAL var_pop((pagg_tab_2.b)::real)), (PARTIAL var_pop((pagg_tab_2.b)::double precision)), (PARTIAL var_pop((pagg_tab_2.b)::smallint)), (PARTIAL var_pop(pagg_tab_2.b)), (PARTIAL var_pop((pagg_tab_2.b)::bigint)), (PARTIAL var_pop((pagg_tab_2.b)::numeric)), (PARTIAL var_samp((pagg_tab_2.b)::real)), (PARTIAL var_samp((pagg_tab_2.b)::double precision)), (PARTIAL var_samp((pagg_tab_2.b)::smallint)), (PARTIAL var_samp(pagg_tab_2.b)), (PARTIAL var_samp((pagg_tab_2.b)::bigint)), (PARTIAL var_samp((pagg_tab_2.b)::numeric)), (PARTIAL any_value((pagg_tab_2.b * 0))), (PARTIAL bit_and(pagg_tab_2.c_bit)), (PARTIAL bit_and(pagg_tab_2.c_1or3int2)), (PARTIAL bit_and(pagg_tab_2.c_1or3int4)), (PARTIAL bit_and(pagg_tab_2.c_1or3int8)), (PARTIAL bit_or(pagg_tab_2.c_bit)), (PARTIAL bit_or(pagg_tab_2.c_1or3int2)), (PARTIAL bit_or(pagg_tab_2.c_1or3int4)), (PARTIAL bit_or(pagg_tab_2.c_1or3int8)), (PARTIAL bit_xor(pagg_tab_2.c_bit)), (PARTIAL bit_xor(pagg_tab_2.c_1or3int2)), (PARTIAL bit_xor(pagg_tab_2.c_1or3int4)), (PARTIAL bit_xor(pagg_tab_2.c_1or3int8)), (PARTIAL bool_and(pagg_tab_2.c_bool)), (PARTIAL bool_or(pagg_tab_2.c_bool)), (PARTIAL count(pagg_tab_2.b)), (PARTIAL count(*)), (PARTIAL every(pagg_tab_2.c_bool)), (PARTIAL max(pagg_tab_2.c_int4array)), (PARTIAL max(pagg_tab_2.c_enum)), (PARTIAL max((pagg_tab_2.c_1c)::character(1))), (PARTIAL max(('01-01-2000'::date + pagg_tab_2.b))), (PARTIAL max(('0.0.0.0'::inet + (pagg_tab_2.b)::bigint))), (PARTIAL max((pagg_tab_2.b)::real)), (PARTIAL max((pagg_tab_2.b)::double precision)), (PARTIAL max((pagg_tab_2.b)::smallint)), (PARTIAL max(pagg_tab_2.b)), (PARTIAL max((pagg_tab_2.b)::bigint)), (PARTIAL max(pagg_tab_2.c_interval)), (PARTIAL max(pagg_tab_2.c_money)), (PARTIAL max((pagg_tab_2.b)::numeric)), (PARTIAL max((pagg_tab_2.b)::oid)), (PARTIAL max(pagg_tab_2.c_pg_lsn)), (PARTIAL max(pagg_tab_2.c_tid)), (PARTIAL max(pagg_tab_2.c_1c)), (PARTIAL max(pagg_tab_2.c_time)), (PARTIAL max(pagg_tab_2.c_timetz)), (PARTIAL max(pagg_tab_2.c_timestamp)), (PARTIAL max(pagg_tab_2.c_timestamptz)), (PARTIAL max(pagg_tab_2.c_xid8)), (PARTIAL min(pagg_tab_2.c_int4array)), (PARTIAL min(pagg_tab_2.c_enum)), (PARTIAL min((pagg_tab_2.c_1c)::character(1))), (PARTIAL min(('01-01-2000'::date + pagg_tab_2.b))), (PARTIAL min(('0.0.0.0'::inet + (pagg_tab_2.b)::bigint))), (PARTIAL min((pagg_tab_2.b)::real)), (PARTIAL min((pagg_tab_2.b)::double precision)), (PARTIAL min((pagg_tab_2.b)::smallint)), (PARTIAL min(pagg_tab_2.b)), (PARTIAL min((pagg_tab_2.b)::bigint)), (PARTIAL min(pagg_tab_2.c_interval)), (PARTIAL min(pagg_tab_2.c_money)), (PARTIAL min((pagg_tab_2.b)::numeric)), (PARTIAL min((pagg_tab_2.b)::oid)), (PARTIAL min(pagg_tab_2.c_pg_lsn)), (PARTIAL min(pagg_tab_2.c_tid)), (PARTIAL min(pagg_tab_2.c_1c)), (PARTIAL min(pagg_tab_2.c_time)), (PARTIAL min(pagg_tab_2.c_timetz)), (PARTIAL min(pagg_tab_2.c_timestamp)), (PARTIAL min(pagg_tab_2.c_timestamptz)), (PARTIAL min(pagg_tab_2.c_xid8)), (PARTIAL range_intersect_agg(pagg_tab_2.c_int4range)), (PARTIAL range_intersect_agg(pagg_tab_2.c_int4multirange)), (PARTIAL regr_count((((2 * pagg_tab_2.b) + 3))::double precision, (pagg_tab_2.b)::double precision)), (PARTIAL sum((pagg_tab_2.b)::real)), (PARTIAL sum((pagg_tab_2.b)::double precision)), (PARTIAL sum((pagg_tab_2.b)::smallint)), (PARTIAL sum(pagg_tab_2.b)), (PARTIAL sum(pagg_tab_2.c_interval)), (PARTIAL sum(pagg_tab_2.c_money)) + Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2) + Remote SQL: SELECT array_agg_p_anyarray(c_int4array), array_agg_p_anynonarray(b), avg_p_int2(b::smallint), avg_p_int4(b), avg_p_int8(b::bigint), avg_p_interval(c_interval), avg_p_float4(b::real), avg_p_float8(b::double precision), avg_p_numeric(b::numeric), corr_p(b::double precision, (b * b)::double precision), covar_pop_p(b::double precision, (b * b)::double precision), covar_samp_p(b::double precision, (b * b)::double precision), regr_avgx_p((2 * b)::double precision, b::double precision), regr_avgy_p((2 * b)::double precision, b::double precision), regr_intercept_p(((2 * b) + 3)::double precision, b::double precision), regr_r2_p((b * b)::double precision, b::double precision), regr_slope_p(((2 * b) + 3)::double precision, b::double precision), regr_sxx_p(((2 * b) + 3)::double precision, b::double precision), regr_sxy_p((b * b)::double precision, b::double precision), regr_syy_p(((2 * b) + 3)::double precision, b::double precision), stddev_p_float4(b::real), stddev_p_float8(b::double precision), stddev_p_int2(b::smallint), stddev_p_int4(b), stddev_p_int8(b::bigint), stddev_p_numeric(b::numeric), stddev_pop_p_float4(b::real), stddev_pop_p_float8(b::double precision), stddev_pop_p_int2(b::smallint), stddev_pop_p_int4(b), stddev_pop_p_int8(b::bigint), stddev_pop_p_numeric(b::numeric), stddev_samp_p_float4(b::real), stddev_samp_p_float8(b::double precision), stddev_samp_p_int2(b::smallint), stddev_samp_p_int4(b), stddev_samp_p_int8(b::bigint), stddev_samp_p_numeric(b::numeric), string_agg_p_text_text(c_1c, ','::text), string_agg_p_bytea_bytea(c_1b, E'\\x2c'::bytea), sum_p_int8(b::bigint), sum_p_numeric(b::numeric), variance_p_float4(b::real), variance_p_float8(b::double precision), variance_p_int2(b::smallint), variance_p_int4(b), variance_p_int8(b::bigint), variance_p_numeric(b::numeric), var_pop_p_float4(b::real), var_pop_p_float8(b::double precision), var_pop_p_int2(b::smallint), var_pop_p_int4(b), var_pop_p_int8(b::bigint), var_pop_p_numeric(b::numeric), var_samp_p_float4(b::real), var_samp_p_float8(b::double precision), var_samp_p_int2(b::smallint), var_samp_p_int4(b), var_samp_p_int8(b::bigint), var_samp_p_numeric(b::numeric), any_value((b * 0)), bit_and(c_bit), bit_and(c_1or3int2), bit_and(c_1or3int4), bit_and(c_1or3int8), bit_or(c_bit), bit_or(c_1or3int2), bit_or(c_1or3int4), bit_or(c_1or3int8), bit_xor(c_bit), bit_xor(c_1or3int2), bit_xor(c_1or3int4), bit_xor(c_1or3int8), bool_and(c_bool), bool_or(c_bool), count(b), count(*), every(c_bool), max(c_int4array), max(c_enum), max(c_1c::character(1)), max(('01-01-2000'::date + b)), max(('0.0.0.0'::inet + b)), max(b::real), max(b::double precision), max(b::smallint), max(b), max(b::bigint), max(c_interval), max(c_money), max(b::numeric), max(b::oid), max(c_pg_lsn), max(c_tid), max(c_1c), max(c_time), max(c_timetz), max(c_timestamp), max(c_timestamptz), max(c_xid8), min(c_int4array), min(c_enum), min(c_1c::character(1)), min(('01-01-2000'::date + b)), min(('0.0.0.0'::inet + b)), min(b::real), min(b::double precision), min(b::smallint), min(b), min(b::bigint), min(c_interval), min(c_money), min(b::numeric), min(b::oid), min(c_pg_lsn), min(c_tid), min(c_1c), min(c_time), min(c_timetz), min(c_timestamp), min(c_timestamptz), min(c_xid8), range_intersect_agg(c_int4range), range_intersect_agg(c_int4multirange), regr_count(((2 * b) + 3)::double precision, b::double precision), sum(b::real), sum(b::double precision), sum(b::smallint), sum(b), sum(c_interval), sum(c_money) FROM public.pagg_tab_p3 WHERE ((c_serial >= 1)) AND ((c_serial <= 30)) +(15 rows) + +SELECT /* aggregate function <> aggpartialfunc */ + array_agg(c_int4array), array_agg(b), + avg(b::int2), avg(b::int4), avg(b::int8), avg(c_interval), + avg(b::float4), avg(b::float8), + corr(b::float8, (b * b)::float8), + covar_pop(b::float8, (b * b)::float8), + covar_samp(b::float8, (b * b)::float8), + regr_avgx((2 * b)::float8, b::float8), + regr_avgy((2 * b)::float8, b::float8), + regr_intercept((2 * b + 3)::float8, b::float8), + regr_r2((b * b)::float8, b::float8), + regr_slope((2 * b + 3)::float8, b::float8), + regr_sxx((2 * b + 3)::float8, b::float8), + regr_sxy((b * b)::float8, b::float8), + regr_syy((2 * b + 3)::float8, b::float8), + stddev(b::float4), stddev(b::float8), stddev(b::int2), stddev(b::int4), stddev(b::int8), stddev(b::numeric), + stddev_pop(b::float4), stddev_pop(b::float8), stddev_pop(b::int2), stddev_pop(b::int4), stddev_pop(b::int8), stddev_pop(b::numeric), + stddev_samp(b::float4), stddev_samp(b::float8), stddev_samp(b::int2), stddev_samp(b::int4), stddev_samp(b::int8), stddev_samp(b::numeric), + string_agg(c_1c, ','), string_agg(c_1b, ','), + sum(b::int8), sum(b::numeric), + variance(b::float4), variance(b::float8), variance(b::int2), variance(b::int4), variance(b::int8), variance(b::numeric), + var_pop(b::float4), var_pop(b::float8), var_pop(b::int2), var_pop(b::int4), var_pop(b::int8), var_pop(b::numeric), + var_samp(b::float4), var_samp(b::float8), var_samp(b::int2), var_samp(b::int4), var_samp(b::int8), var_samp(b::numeric), + /* aggregate function = aggpartialfunc */ + any_value(b * 0), + bit_and(c_bit), bit_and(c_1or3int2), bit_and(c_1or3int4), bit_and(c_1or3int8), + bit_or(c_bit), bit_or(c_1or3int2), bit_or(c_1or3int4), bit_or(c_1or3int8), + bit_xor(c_bit), bit_xor(c_1or3int2), bit_xor(c_1or3int4), bit_xor(c_1or3int8), + bool_and(c_bool), + bool_or(c_bool), + count(b), count(*), + every(c_bool), + max(c_int4array), max(c_enum), max(c_1c::char(1)), max('2000-01-01'::date + b), max('0.0.0.0'::inet + b), max(b::float4), max(b::float8), max(b::int2), max(b::int4), max(b::int8), max(c_interval), max(c_money), max(b::numeric), max(b::oid), max(c_pg_lsn), max(c_tid), max(c_1c), max(c_time), max(c_timetz), max(c_timestamp), max(c_timestamptz), max(c_xid8), + min(c_int4array), min(c_enum), min(c_1c::char(1)), min('2000-01-01'::date + b), min('0.0.0.0'::inet + b), min(b::float4), min(b::float8), min(b::int2), min(b::int4), min(b::int8), min(c_interval), min(c_money), min(b::numeric), min(b::oid), min(c_pg_lsn), min(c_tid), min(c_1c), min(c_time), min(c_timetz), min(c_timestamp), min(c_timestamptz), min(c_xid8), + range_intersect_agg(c_int4range), range_intersect_agg(c_int4multirange), + regr_count((2 * b + 3)::float8, b::float8), + sum(b::float4), sum(b::float8), sum(b::int2), sum(b::int4), sum(c_interval), sum(c_money) + FROM pagg_tab WHERE c_serial between 1 and 30; + array_agg | array_agg | avg | avg | avg | avg | avg | avg | corr | covar_pop | covar_samp | regr_avgx | regr_avgy | regr_intercept | regr_r2 | regr_slope | regr_sxx | regr_sxy | regr_syy | stddev | stddev | stddev | stddev | stddev | stddev | stddev_pop | stddev_pop | stddev_pop | stddev_pop | stddev_pop | stddev_pop | stddev_samp | stddev_samp | stddev_samp | stddev_samp | stddev_samp | stddev_samp | string_agg | string_agg | sum | sum | variance | variance | variance | variance | variance | variance | var_pop | var_pop | var_pop | var_pop | var_pop | var_pop | var_samp | var_samp | var_samp | var_samp | var_samp | var_samp | any_value | bit_and | bit_and | bit_and | bit_and | bit_or | bit_or | bit_or | bit_or | bit_xor | bit_xor | bit_xor | bit_xor | bool_and | bool_or | count | count | every | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | min | min | min | min | min | min | min | min | min | min | min | min | min | min | min | min | min | min | min | min | min | min | range_intersect_agg | range_intersect_agg | regr_count | sum | sum | sum | sum | sum | sum +---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+------------------------------------------------------------------------------------+---------------------+---------------------+---------------------+------------+------+------+--------------------+--------------------+------------+-----------+-----------+----------------+--------------------+------------+----------+----------+----------+-------------------+-------------------+--------------------+--------------------+--------------------+--------------------+------------------+------------------+--------------------+--------------------+--------------------+--------------------+-------------------+-------------------+--------------------+--------------------+--------------------+--------------------+-------------------------------------------------------------+--------------------------------------------------------------------------------------------------------------------------+-----+-----+----------+----------+---------------------+---------------------+---------------------+---------------------+-------------------+-------------------+---------------------+---------------------+---------------------+---------------------+----------+----------+---------------------+---------------------+---------------------+---------------------+-----------+---------+---------+---------+---------+--------+--------+--------+--------+---------+---------+---------+---------+----------+---------+-------+-------+-------+-------+-------+-----+------------+----------+-----+-----+-----+-----+-----+---------+-------+-----+-----+------+--------+-----+----------+-------------+--------------------------+------------------------------+-----+-------+-----+-----+------------+---------+-----+-----+-----+-----+-----+-----+-------+-----+-----+-----+-------+-----+----------+-------------+--------------------------+------------------------------+-----+---------------------+---------------------+------------+-----+-----+-----+-----+-----------+-------- + {{1,0},{0,0},{1,0},{0,0},{1,0},{0,0},{1,0},{0,0},{1,0},{0,0},{0,0},{1,0},{0,0},{1,0},{0,0},{1,0},{0,0},{1,0},{0,0},{1,0},{0,0},{1,0},{0,0},{1,0},{0,0},{1,0},{0,0},{1,0},{0,0},{1,0}} | {1,2,3,4,5,6,7,8,9,30,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29} | 15.5000000000000000 | 15.5000000000000000 | 15.5000000000000000 | @ 0.5 secs | 15.5 | 15.5 | 0.9702989135892578 | 2322.4166666666665 | 2402.5 | 15.5 | 31 | 3 | 0.9414799817124941 | 2 | 2247.5 | 69672.5 | 8990 | 8.803408430829505 | 8.803408430829505 | 8.8034084308295046 | 8.8034084308295046 | 8.8034084308295046 | 8.8034084308295046 | 8.65544144839919 | 8.65544144839919 | 8.6554414483991899 | 8.6554414483991899 | 8.6554414483991899 | 8.6554414483991899 | 8.803408430829505 | 8.803408430829505 | 8.8034084308295046 | 8.8034084308295046 | 8.8034084308295046 | 8.8034084308295046 | 0,1,2,3,4,5,6,7,8,9,9,0,1,2,3,4,5,6,7,8,9,0,1,2,3,4,5,6,7,8 | \x302c312c322c332c342c352c362c372c382c392c392c302c312c322c332c342c352c362c372c382c392c302c312c322c332c342c352c362c372c38 | 465 | 465 | 77.5 | 77.5 | 77.5000000000000000 | 77.5000000000000000 | 77.5000000000000000 | 77.5000000000000000 | 74.91666666666667 | 74.91666666666667 | 74.9166666666666667 | 74.9166666666666667 | 74.9166666666666667 | 74.9166666666666667 | 77.5 | 77.5 | 77.5000000000000000 | 77.5000000000000000 | 77.5000000000000000 | 77.5000000000000000 | 0 | 01 | 1 | 1 | 1 | 11 | 3 | 3 | 3 | 10 | 2 | 2 | 2 | f | t | 30 | 30 | f | {1,0} | happy | 9 | 01-31-2000 | 0.0.0.30 | 30 | 30 | 30 | 30 | 30 | @ 1 sec | $1.00 | 30 | 30 | 0/30 | (0,30) | 9 | 00:00:30 | 00:00:30-07 | Sat Jan 01 00:00:30 2000 | Sat Jan 01 00:00:30 2000 PST | 9 | {0,0} | sad | 0 | 01-02-2000 | 0.0.0.1 | 1 | 1 | 1 | 1 | 1 | @ 0 | $0.00 | 1 | 1 | 0/1 | (0,1) | 0 | 00:00:01 | 00:00:01-07 | Sat Jan 01 00:00:01 2000 | Sat Jan 01 00:00:01 2000 PST | 0 | [0,1) | {[0,1),[100,101)} | 30 | 465 | 465 | 465 | 465 | @ 15 secs | $15.00 +(1 row) + +-- Tests for backward compatibility +ALTER SERVER loopback OPTIONS (ADD check_partial_aggregate_support 'false'); +EXPLAIN (VERBOSE, COSTS OFF) +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; + QUERY PLAN +------------------------------------------------------------------------------------------------------------------------ + Sort + Output: pagg_tab.b, (avg(pagg_tab.a)), (max(pagg_tab.a)), (count(*)) + Sort Key: pagg_tab.b + -> Finalize HashAggregate + Output: pagg_tab.b, avg(pagg_tab.a), max(pagg_tab.a), count(*) + Group Key: pagg_tab.b + -> Append + -> Foreign Scan + Output: pagg_tab.b, (PARTIAL avg(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab) + Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p1 GROUP BY 1 + -> Foreign Scan + Output: pagg_tab_1.b, (PARTIAL avg(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1) + Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p2 GROUP BY 1 + -> Foreign Scan + Output: pagg_tab_2.b, (PARTIAL avg(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2) + Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p3 GROUP BY 1 +(19 rows) + +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; + b | avg | max | count +----+---------------------+-----+------- + 0 | 10.0000000000000000 | 20 | 60 + 1 | 11.0000000000000000 | 21 | 60 + 2 | 12.0000000000000000 | 22 | 60 + 3 | 13.0000000000000000 | 23 | 60 + 4 | 14.0000000000000000 | 24 | 60 + 5 | 15.0000000000000000 | 25 | 60 + 6 | 16.0000000000000000 | 26 | 60 + 7 | 17.0000000000000000 | 27 | 60 + 8 | 18.0000000000000000 | 28 | 60 + 9 | 19.0000000000000000 | 29 | 60 + 10 | 10.0000000000000000 | 20 | 60 + 11 | 11.0000000000000000 | 21 | 60 + 12 | 12.0000000000000000 | 22 | 60 + 13 | 13.0000000000000000 | 23 | 60 + 14 | 14.0000000000000000 | 24 | 60 + 15 | 15.0000000000000000 | 25 | 60 + 16 | 16.0000000000000000 | 26 | 60 + 17 | 17.0000000000000000 | 27 | 60 + 18 | 18.0000000000000000 | 28 | 60 + 19 | 19.0000000000000000 | 29 | 60 + 20 | 10.0000000000000000 | 20 | 60 + 21 | 11.0000000000000000 | 21 | 60 + 22 | 12.0000000000000000 | 22 | 60 + 23 | 13.0000000000000000 | 23 | 60 + 24 | 14.0000000000000000 | 24 | 60 + 25 | 15.0000000000000000 | 25 | 60 + 26 | 16.0000000000000000 | 26 | 60 + 27 | 17.0000000000000000 | 27 | 60 + 28 | 18.0000000000000000 | 28 | 60 + 29 | 19.0000000000000000 | 29 | 60 + 30 | 10.0000000000000000 | 20 | 60 + 31 | 11.0000000000000000 | 21 | 60 + 32 | 12.0000000000000000 | 22 | 60 + 33 | 13.0000000000000000 | 23 | 60 + 34 | 14.0000000000000000 | 24 | 60 + 35 | 15.0000000000000000 | 25 | 60 + 36 | 16.0000000000000000 | 26 | 60 + 37 | 17.0000000000000000 | 27 | 60 + 38 | 18.0000000000000000 | 28 | 60 + 39 | 19.0000000000000000 | 29 | 60 + 40 | 10.0000000000000000 | 20 | 60 + 41 | 11.0000000000000000 | 21 | 60 + 42 | 12.0000000000000000 | 22 | 60 + 43 | 13.0000000000000000 | 23 | 60 + 44 | 14.0000000000000000 | 24 | 60 + 45 | 15.0000000000000000 | 25 | 60 + 46 | 16.0000000000000000 | 26 | 60 + 47 | 17.0000000000000000 | 27 | 60 + 48 | 18.0000000000000000 | 28 | 60 + 49 | 19.0000000000000000 | 29 | 60 +(50 rows) + +ALTER SERVER loopback OPTIONS (SET check_partial_aggregate_support 'true'); +EXPLAIN (VERBOSE, COSTS OFF) +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; + QUERY PLAN +------------------------------------------------------------------------------------------------------------------------ + Sort + Output: pagg_tab.b, (avg(pagg_tab.a)), (max(pagg_tab.a)), (count(*)) + Sort Key: pagg_tab.b + -> Finalize HashAggregate + Output: pagg_tab.b, avg(pagg_tab.a), max(pagg_tab.a), count(*) + Group Key: pagg_tab.b + -> Append + -> Foreign Scan + Output: pagg_tab.b, (PARTIAL avg(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab) + Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p1 GROUP BY 1 + -> Foreign Scan + Output: pagg_tab_1.b, (PARTIAL avg(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1) + Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p2 GROUP BY 1 + -> Foreign Scan + Output: pagg_tab_2.b, (PARTIAL avg(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)) + Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2) + Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*) FROM public.pagg_tab_p3 GROUP BY 1 +(19 rows) + +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; + b | avg | max | count +----+---------------------+-----+------- + 0 | 10.0000000000000000 | 20 | 60 + 1 | 11.0000000000000000 | 21 | 60 + 2 | 12.0000000000000000 | 22 | 60 + 3 | 13.0000000000000000 | 23 | 60 + 4 | 14.0000000000000000 | 24 | 60 + 5 | 15.0000000000000000 | 25 | 60 + 6 | 16.0000000000000000 | 26 | 60 + 7 | 17.0000000000000000 | 27 | 60 + 8 | 18.0000000000000000 | 28 | 60 + 9 | 19.0000000000000000 | 29 | 60 + 10 | 10.0000000000000000 | 20 | 60 + 11 | 11.0000000000000000 | 21 | 60 + 12 | 12.0000000000000000 | 22 | 60 + 13 | 13.0000000000000000 | 23 | 60 + 14 | 14.0000000000000000 | 24 | 60 + 15 | 15.0000000000000000 | 25 | 60 + 16 | 16.0000000000000000 | 26 | 60 + 17 | 17.0000000000000000 | 27 | 60 + 18 | 18.0000000000000000 | 28 | 60 + 19 | 19.0000000000000000 | 29 | 60 + 20 | 10.0000000000000000 | 20 | 60 + 21 | 11.0000000000000000 | 21 | 60 + 22 | 12.0000000000000000 | 22 | 60 + 23 | 13.0000000000000000 | 23 | 60 + 24 | 14.0000000000000000 | 24 | 60 + 25 | 15.0000000000000000 | 25 | 60 + 26 | 16.0000000000000000 | 26 | 60 + 27 | 17.0000000000000000 | 27 | 60 + 28 | 18.0000000000000000 | 28 | 60 + 29 | 19.0000000000000000 | 29 | 60 + 30 | 10.0000000000000000 | 20 | 60 + 31 | 11.0000000000000000 | 21 | 60 + 32 | 12.0000000000000000 | 22 | 60 + 33 | 13.0000000000000000 | 23 | 60 + 34 | 14.0000000000000000 | 24 | 60 + 35 | 15.0000000000000000 | 25 | 60 + 36 | 16.0000000000000000 | 26 | 60 + 37 | 17.0000000000000000 | 27 | 60 + 38 | 18.0000000000000000 | 28 | 60 + 39 | 19.0000000000000000 | 29 | 60 + 40 | 10.0000000000000000 | 20 | 60 + 41 | 11.0000000000000000 | 21 | 60 + 42 | 12.0000000000000000 | 22 | 60 + 43 | 13.0000000000000000 | 23 | 60 + 44 | 14.0000000000000000 | 24 | 60 + 45 | 15.0000000000000000 | 25 | 60 + 46 | 16.0000000000000000 | 26 | 60 + 47 | 17.0000000000000000 | 27 | 60 + 48 | 18.0000000000000000 | 28 | 60 + 49 | 19.0000000000000000 | 29 | 60 +(50 rows) + +ALTER SERVER loopback OPTIONS (SET fdw_tuple_cost '1.0'); +SET enable_partitionwise_join=on; +EXPLAIN (VERBOSE, COSTS OFF) +SELECT avg(t1.b), avg(t1.b::int8) FROM pagg_tab t1 JOIN pagg_tab t2 USING(a); + QUERY PLAN +---------------------------------------------------------------------------------------------------------------------------------------------------------------- + Finalize Aggregate + Output: avg(t1.b), avg((t1.b)::bigint) + -> Append + -> Foreign Scan + Output: (PARTIAL avg(t1.b)), (PARTIAL avg((t1.b)::bigint)) + Relations: Aggregate on ((public.fpagg_tab_p1 t1) INNER JOIN (public.fpagg_tab_p1 t2)) + Remote SQL: SELECT avg_p_int4(r4.b), avg_p_int8(r4.b::bigint) FROM (public.pagg_tab_p1 r4 INNER JOIN public.pagg_tab_p1 r7 ON (((r4.a = r7.a)))) + -> Foreign Scan + Output: (PARTIAL avg(t1_1.b)), (PARTIAL avg((t1_1.b)::bigint)) + Relations: Aggregate on ((public.fpagg_tab_p2 t1_1) INNER JOIN (public.fpagg_tab_p2 t2_1)) + Remote SQL: SELECT avg_p_int4(r5.b), avg_p_int8(r5.b::bigint) FROM (public.pagg_tab_p2 r5 INNER JOIN public.pagg_tab_p2 r8 ON (((r5.a = r8.a)))) + -> Foreign Scan + Output: (PARTIAL avg(t1_2.b)), (PARTIAL avg((t1_2.b)::bigint)) + Relations: Aggregate on ((public.fpagg_tab_p3 t1_2) INNER JOIN (public.fpagg_tab_p3 t2_2)) + Remote SQL: SELECT avg_p_int4(r6.b), avg_p_int8(r6.b::bigint) FROM (public.pagg_tab_p3 r6 INNER JOIN public.pagg_tab_p3 r9 ON (((r6.a = r9.a)))) +(15 rows) + +SELECT avg(t1.b), avg(t1.b::int8) FROM pagg_tab t1 JOIN pagg_tab t2 USING(a); + avg | avg +---------------------+--------------------- + 24.5000000000000000 | 24.5000000000000000 +(1 row) + +RESET enable_partitionwise_join; +ALTER SERVER loopback OPTIONS (SET fdw_tuple_cost '0.1'); +ALTER SERVER loopback OPTIONS (DROP check_partial_aggregate_support); +-- List of aggregate functions such that the aggregate function supports partial aggregate +-- and aggpartialfn field of the aggregate function is invalid +SELECT aggfnoid::text + FROM pg_aggregate a JOIN pg_type b ON a.aggtranstype = b.oid + WHERE (aggcombinefn::text <> '-') AND (aggpartialfn::text = '-') + ORDER BY 1; + aggfnoid +---------- +(0 rows) + +-- List of aggregate functions which have incorrect aggpartialfunc +SELECT a.aggfnoid + FROM pg_aggregate a JOIN pg_type b ON a.aggtranstype = b.oid + JOIN pg_aggregate c on a.aggpartialfn = c.aggfnoid + WHERE (a.aggcombinefn::text <> '-') + AND (a.aggpartialfn::text <> '-') + AND (a.aggtransfn <> c.aggtransfn + OR (c.aggcombinefn::text <> '-' + AND a.aggcombinefn <> c.aggcombinefn) + OR (a.agginitval IS NOT NULL + AND c.agginitval IS NOT NULL + AND a.agginitval <> c.agginitval) + OR (b.typname = 'internal' + AND c.aggserialfn::text <> '-' + AND a.aggserialfn <> c.aggserialfn) + OR (b.typname = 'internal' + AND c.aggdeserialfn::text <> '-' + AND a.aggdeserialfn <> c.aggdeserialfn) + OR (b.typname = 'internal' + AND a.aggserialfn <> c.aggfinalfn) + OR (b.typname <> 'internal' + AND c.aggfinalfn::text <> '-')) + ORDER BY 1; + aggfnoid +---------- +(0 rows) + +-- It's unsafe to push down partial aggregates which contain distinct clause +EXPLAIN (VERBOSE, COSTS OFF) +SELECT max(a), count(distinct b) FROM pagg_tab; + QUERY PLAN +----------------------------------------------------------------------------------------- + Aggregate + Output: max(pagg_tab.a), count(DISTINCT pagg_tab.b) + -> Merge Append + Sort Key: pagg_tab.b + -> Foreign Scan on public.fpagg_tab_p1 pagg_tab_1 + Output: pagg_tab_1.a, pagg_tab_1.b + Remote SQL: SELECT a, b FROM public.pagg_tab_p1 ORDER BY b ASC NULLS LAST + -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_2 + Output: pagg_tab_2.a, pagg_tab_2.b + Remote SQL: SELECT a, b FROM public.pagg_tab_p2 ORDER BY b ASC NULLS LAST + -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_3 + Output: pagg_tab_3.a, pagg_tab_3.b + Remote SQL: SELECT a, b FROM public.pagg_tab_p3 ORDER BY b ASC NULLS LAST +(13 rows) + +SELECT max(a), count(distinct b) FROM pagg_tab; + max | count +-----+------- + 29 | 50 +(1 row) + +-- It's unsafe to push down partial aggregates which contain order by clause +EXPLAIN (VERBOSE, COSTS OFF) SELECT array_agg(b order by b) FROM pagg_tab WHERE c_serial between 1 and 30; + QUERY PLAN +----------------------------------------------------------------------------------------------------------------- + Aggregate + Output: array_agg(pagg_tab.b ORDER BY pagg_tab.b) + -> Sort + Output: pagg_tab.b + Sort Key: pagg_tab.b + -> Append + -> Foreign Scan on public.fpagg_tab_p1 pagg_tab_1 + Output: pagg_tab_1.b + Remote SQL: SELECT b FROM public.pagg_tab_p1 WHERE ((c_serial >= 1)) AND ((c_serial <= 30)) + -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_2 + Output: pagg_tab_2.b + Remote SQL: SELECT b FROM public.pagg_tab_p2 WHERE ((c_serial >= 1)) AND ((c_serial <= 30)) + -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_3 + Output: pagg_tab_3.b + Remote SQL: SELECT b FROM public.pagg_tab_p3 WHERE ((c_serial >= 1)) AND ((c_serial <= 30)) (15 rows) +SELECT array_agg(b order by b) FROM pagg_tab WHERE c_serial between 1 and 30; + array_agg +------------------------------------------------------------------------------------ + {1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30} +(1 row) + +-- Create user-defined aggregates whose stype is internal +CREATE SCHEMA postgres_fdw_test; +CREATE AGGREGATE postgres_fdw_test.postgres_fdw_sum_p_int8(int8) ( + SFUNC = int8_avg_accum, + STYPE = internal, + COMBINEFUNC = int8_avg_combine, + FINALFUNC = int8_avg_serialize, + SERIALFUNC = int8_avg_serialize, + DESERIALFUNC = int8_avg_deserialize, + AGGPARTIALFUNC = postgres_fdw_test.postgres_fdw_sum_p_int8 +); +CREATE AGGREGATE postgres_fdw_sum(int8) ( + SFUNC = int8_avg_accum, + STYPE = internal, + COMBINEFUNC = int8_avg_combine, + FINALFUNC = numeric_poly_sum, + SERIALFUNC = int8_avg_serialize, + DESERIALFUNC = int8_avg_deserialize, + AGGPARTIALFUNC = postgres_fdw_test.postgres_fdw_sum_p_int8 +); +-- Create user-defined aggregates whose stype is not internal +CREATE AGGREGATE postgres_fdw_avg_p_int4(int4) ( + SFUNC = int4_avg_accum, + STYPE = _int8, + COMBINEFUNC = int4_avg_combine, + INITCOND = '{0,0}', + AGGPARTIALFUNC = postgres_fdw_avg_p_int4 +); +CREATE AGGREGATE postgres_fdw_avg(int4) ( + SFUNC = int4_avg_accum, + STYPE = _int8, + COMBINEFUNC = int4_avg_combine, + FINALFUNC = int8_avg, + INITCOND = '{0,0}', + AGGPARTIALFUNC = postgres_fdw_avg_p_int4 +); +CREATE AGGREGATE postgres_fdw_sum_noaggpartialfunc(int8) ( + SFUNC = int8_avg_accum, + STYPE = internal, + COMBINEFUNC = int8_avg_combine, + FINALFUNC = numeric_poly_sum, + SERIALFUNC = int8_avg_serialize, + DESERIALFUNC = int8_avg_deserialize +); +-- It's safe to push down partial aggregate for user-defined aggregate +-- when it belongs to an extension that's listed in extensions set +-- and when its aggpartialfunc belongs to an extension that's listed +-- in extensions set. +ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_sum(int8); +ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_test.postgres_fdw_sum_p_int8(int8); +ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_avg(int4); +ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_avg_p_int4(int4); +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_sum(b::int8) FROM pagg_tab; + QUERY PLAN +--------------------------------------------------------------------------------------------------------------- + Finalize Aggregate + Output: postgres_fdw_sum((pagg_tab.b)::bigint) + -> Append + -> Foreign Scan + Output: (PARTIAL postgres_fdw_sum((pagg_tab.b)::bigint)) + Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab) + Remote SQL: SELECT postgres_fdw_test.postgres_fdw_sum_p_int8(b::bigint) FROM public.pagg_tab_p1 + -> Foreign Scan + Output: (PARTIAL postgres_fdw_sum((pagg_tab_1.b)::bigint)) + Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1) + Remote SQL: SELECT postgres_fdw_test.postgres_fdw_sum_p_int8(b::bigint) FROM public.pagg_tab_p2 + -> Foreign Scan + Output: (PARTIAL postgres_fdw_sum((pagg_tab_2.b)::bigint)) + Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2) + Remote SQL: SELECT postgres_fdw_test.postgres_fdw_sum_p_int8(b::bigint) FROM public.pagg_tab_p3 +(15 rows) + +SELECT postgres_fdw_sum(b::int8) FROM pagg_tab; + postgres_fdw_sum +------------------ + 73500 +(1 row) + +-- It's unsafe to push down partial aggregate for user-defined aggregate +-- when it has no aggpartialfunc. +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_sum_noaggpartialfunc(b::int8) FROM pagg_tab; + QUERY PLAN +----------------------------------------------------------------------------------------- + Finalize Aggregate + Output: postgres_fdw_sum_noaggpartialfunc((pagg_tab.b)::bigint) + -> Append + -> Partial Aggregate + Output: PARTIAL postgres_fdw_sum_noaggpartialfunc((pagg_tab.b)::bigint) + -> Foreign Scan on public.fpagg_tab_p1 pagg_tab + Output: pagg_tab.b + Remote SQL: SELECT b FROM public.pagg_tab_p1 + -> Partial Aggregate + Output: PARTIAL postgres_fdw_sum_noaggpartialfunc((pagg_tab_1.b)::bigint) + -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1 + Output: pagg_tab_1.b + Remote SQL: SELECT b FROM public.pagg_tab_p2 + -> Partial Aggregate + Output: PARTIAL postgres_fdw_sum_noaggpartialfunc((pagg_tab_2.b)::bigint) + -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2 + Output: pagg_tab_2.b + Remote SQL: SELECT b FROM public.pagg_tab_p3 +(18 rows) + +SELECT postgres_fdw_sum_noaggpartialfunc(b::int8) FROM pagg_tab; + postgres_fdw_sum_noaggpartialfunc +----------------------------------- + 73500 +(1 row) + +ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_test.postgres_fdw_sum_p_int8(int8); +-- Reconnect for flushing the shippability cache +\c - +SET enable_partitionwise_aggregate TO true; +-- It's unsafe to push down partial aggregate for user-defined aggregate +-- when itself is aggpartialfunc and when it belongs to an extension +-- that's listed in extensions set and when stype is internal. +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_test.postgres_fdw_sum_p_int8(b::int8) FROM pagg_tab; + QUERY PLAN +------------------------------------------------------------------------------------------------- + Finalize Aggregate + Output: postgres_fdw_test.postgres_fdw_sum_p_int8((pagg_tab.b)::bigint) + -> Append + -> Partial Aggregate + Output: PARTIAL postgres_fdw_test.postgres_fdw_sum_p_int8((pagg_tab.b)::bigint) + -> Foreign Scan on public.fpagg_tab_p1 pagg_tab + Output: pagg_tab.b + Remote SQL: SELECT b FROM public.pagg_tab_p1 + -> Partial Aggregate + Output: PARTIAL postgres_fdw_test.postgres_fdw_sum_p_int8((pagg_tab_1.b)::bigint) + -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1 + Output: pagg_tab_1.b + Remote SQL: SELECT b FROM public.pagg_tab_p2 + -> Partial Aggregate + Output: PARTIAL postgres_fdw_test.postgres_fdw_sum_p_int8((pagg_tab_2.b)::bigint) + -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2 + Output: pagg_tab_2.b + Remote SQL: SELECT b FROM public.pagg_tab_p3 +(18 rows) + +SELECT postgres_fdw_test.postgres_fdw_sum_p_int8(b::int8) FROM pagg_tab; + postgres_fdw_sum_p_int8 +------------------------------------------------------------ + \x0000000000000bb80000000200000001000000000000000000070dac +(1 row) + +ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_sum(int8); +ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_test.postgres_fdw_sum_p_int8(int8); +-- Reconnect for flushing the shippability cache +\c - +SET enable_partitionwise_aggregate TO true; +-- It's unsafe to push down partial aggregate for user-defined aggregate +-- when it doesn't belong to an extension that's listed in extensions set +-- and when its aggpartialfunc belongs to an extension that's listed +-- in extensions set. +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_sum(b::int8) FROM pagg_tab; + QUERY PLAN +------------------------------------------------------------------------ + Finalize Aggregate + Output: postgres_fdw_sum((pagg_tab.b)::bigint) + -> Append + -> Partial Aggregate + Output: PARTIAL postgres_fdw_sum((pagg_tab.b)::bigint) + -> Foreign Scan on public.fpagg_tab_p1 pagg_tab + Output: pagg_tab.b + Remote SQL: SELECT b FROM public.pagg_tab_p1 + -> Partial Aggregate + Output: PARTIAL postgres_fdw_sum((pagg_tab_1.b)::bigint) + -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1 + Output: pagg_tab_1.b + Remote SQL: SELECT b FROM public.pagg_tab_p2 + -> Partial Aggregate + Output: PARTIAL postgres_fdw_sum((pagg_tab_2.b)::bigint) + -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2 + Output: pagg_tab_2.b + Remote SQL: SELECT b FROM public.pagg_tab_p3 +(18 rows) + +SELECT postgres_fdw_sum(b::int8) FROM pagg_tab; + postgres_fdw_sum +------------------ + 73500 +(1 row) + +-- It's safe to push down partial aggregate for user-defined aggregate +-- whose stype is not internal +-- when it belongs to an extension that's listed in extensions set +-- and when its aggpartialfunc belongs to an extension that's listed +-- in extensions set. +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_avg(b) FROM pagg_tab; + QUERY PLAN +-------------------------------------------------------------------------------------------- + Finalize Aggregate + Output: postgres_fdw_avg(pagg_tab.b) + -> Append + -> Foreign Scan + Output: (PARTIAL postgres_fdw_avg(pagg_tab.b)) + Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab) + Remote SQL: SELECT public.postgres_fdw_avg_p_int4(b) FROM public.pagg_tab_p1 + -> Foreign Scan + Output: (PARTIAL postgres_fdw_avg(pagg_tab_1.b)) + Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1) + Remote SQL: SELECT public.postgres_fdw_avg_p_int4(b) FROM public.pagg_tab_p2 + -> Foreign Scan + Output: (PARTIAL postgres_fdw_avg(pagg_tab_2.b)) + Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2) + Remote SQL: SELECT public.postgres_fdw_avg_p_int4(b) FROM public.pagg_tab_p3 +(15 rows) + +SELECT postgres_fdw_avg(b) FROM pagg_tab; + postgres_fdw_avg +--------------------- + 24.5000000000000000 +(1 row) + +-- It's safe to push down partial aggregate for user-defined aggregate +-- when itself is aggpartialfunc and when it belongs to an extension +-- that's listed in extensions set and when stype is not internal. +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_avg_p_int4(b) FROM pagg_tab; + QUERY PLAN +-------------------------------------------------------------------------------------------- + Finalize Aggregate + Output: postgres_fdw_avg_p_int4(pagg_tab.b) + -> Append + -> Foreign Scan + Output: (PARTIAL postgres_fdw_avg_p_int4(pagg_tab.b)) + Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab) + Remote SQL: SELECT public.postgres_fdw_avg_p_int4(b) FROM public.pagg_tab_p1 + -> Foreign Scan + Output: (PARTIAL postgres_fdw_avg_p_int4(pagg_tab_1.b)) + Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1) + Remote SQL: SELECT public.postgres_fdw_avg_p_int4(b) FROM public.pagg_tab_p2 + -> Foreign Scan + Output: (PARTIAL postgres_fdw_avg_p_int4(pagg_tab_2.b)) + Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2) + Remote SQL: SELECT public.postgres_fdw_avg_p_int4(b) FROM public.pagg_tab_p3 +(15 rows) + +SELECT postgres_fdw_avg_p_int4(b) FROM pagg_tab; + postgres_fdw_avg_p_int4 +------------------------- + {3000,73500} +(1 row) + +ALTER EXTENSION postgres_fdw DROP TYPE mood; +ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_test.postgres_fdw_sum_p_int8(int8); +ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_avg_p_int4(int4); +ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_avg(int4); +DROP AGGREGATE postgres_fdw_sum(int8); +DROP AGGREGATE postgres_fdw_test.postgres_fdw_sum_p_int8(int8); +DROP SCHEMA postgres_fdw_test; +DROP AGGREGATE postgres_fdw_avg(int4); +DROP AGGREGATE postgres_fdw_avg_p_int4(int4); +DROP AGGREGATE postgres_fdw_sum_noaggpartialfunc(int8); +ALTER SERVER loopback OPTIONS (DROP fdw_tuple_cost); -- =================================================================== -- access rights and superuser -- =================================================================== diff --git a/contrib/postgres_fdw/option.c b/contrib/postgres_fdw/option.c index 8c822f4ef9..67e15c8bb9 100644 --- a/contrib/postgres_fdw/option.c +++ b/contrib/postgres_fdw/option.c @@ -126,7 +126,8 @@ postgres_fdw_validator(PG_FUNCTION_ARGS) strcmp(def->defname, "async_capable") == 0 || strcmp(def->defname, "parallel_commit") == 0 || strcmp(def->defname, "parallel_abort") == 0 || - strcmp(def->defname, "keep_connections") == 0) + strcmp(def->defname, "keep_connections") == 0 || + strcmp(def->defname, "check_partial_aggregate_support") == 0) { /* these accept only boolean values */ (void) defGetBoolean(def); @@ -268,6 +269,7 @@ InitPgFdwOptions(void) /* batch_size is available on both server and table */ {"batch_size", ForeignServerRelationId, false}, {"batch_size", ForeignTableRelationId, false}, + {"check_partial_aggregate_support", ForeignServerRelationId, false}, /* async_capable is available on both server and table */ {"async_capable", ForeignServerRelationId, false}, {"async_capable", ForeignTableRelationId, false}, diff --git a/contrib/postgres_fdw/postgres_fdw.c b/contrib/postgres_fdw/postgres_fdw.c index c5cada55fb..de71a1600e 100644 --- a/contrib/postgres_fdw/postgres_fdw.c +++ b/contrib/postgres_fdw/postgres_fdw.c @@ -36,6 +36,7 @@ #include "optimizer/pathnode.h" #include "optimizer/paths.h" #include "optimizer/planmain.h" +#include "optimizer/planner.h" #include "optimizer/prep.h" #include "optimizer/restrictinfo.h" #include "optimizer/tlist.h" @@ -519,7 +520,7 @@ static bool foreign_join_ok(PlannerInfo *root, RelOptInfo *joinrel, JoinType jointype, RelOptInfo *outerrel, RelOptInfo *innerrel, JoinPathExtraData *extra); static bool foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel, - Node *havingQual); + GroupPathExtraData *extra); static List *get_useful_pathkeys_for_relation(PlannerInfo *root, RelOptInfo *rel); static List *get_useful_ecs_for_relation(PlannerInfo *root, RelOptInfo *rel); @@ -650,6 +651,8 @@ postgresGetForeignRelSize(PlannerInfo *root, fpinfo->fdw_tuple_cost = DEFAULT_FDW_TUPLE_COST; fpinfo->shippable_extensions = NIL; fpinfo->fetch_size = 100; + fpinfo->check_partial_aggregate_support = false; + fpinfo->remoteversion = 0; fpinfo->async_capable = false; apply_server_options(fpinfo); @@ -661,7 +664,7 @@ postgresGetForeignRelSize(PlannerInfo *root, * should match what ExecCheckPermissions() does. If we fail due to lack * of permissions, the query would have failed at runtime anyway. */ - if (fpinfo->use_remote_estimate) + if (fpinfo->use_remote_estimate || fpinfo->check_partial_aggregate_support) { Oid userid; @@ -5946,9 +5949,10 @@ foreign_join_ok(PlannerInfo *root, RelOptInfo *joinrel, JoinType jointype, fpinfo->pushdown_safe = true; /* Get user mapping */ - if (fpinfo->use_remote_estimate) + if (fpinfo->use_remote_estimate || fpinfo->check_partial_aggregate_support) { - if (fpinfo_o->use_remote_estimate) + if (fpinfo_o->use_remote_estimate || + fpinfo_o->check_partial_aggregate_support) fpinfo->user = fpinfo_o->user; else fpinfo->user = fpinfo_i->user; @@ -6127,6 +6131,8 @@ apply_server_options(PgFdwRelationInfo *fpinfo) ExtractExtensionList(defGetString(def), false); else if (strcmp(def->defname, "fetch_size") == 0) (void) parse_int(defGetString(def), &fpinfo->fetch_size, 0, NULL); + else if (strcmp(def->defname, "check_partial_aggregate_support") == 0) + fpinfo->check_partial_aggregate_support = defGetBoolean(def); else if (strcmp(def->defname, "async_capable") == 0) fpinfo->async_capable = defGetBoolean(def); } @@ -6186,6 +6192,8 @@ merge_fdw_options(PgFdwRelationInfo *fpinfo, fpinfo->shippable_extensions = fpinfo_o->shippable_extensions; fpinfo->use_remote_estimate = fpinfo_o->use_remote_estimate; fpinfo->fetch_size = fpinfo_o->fetch_size; + fpinfo->check_partial_aggregate_support = fpinfo_o->check_partial_aggregate_support; + fpinfo->remoteversion = fpinfo_o->remoteversion; fpinfo->async_capable = fpinfo_o->async_capable; /* Merge the table level options from either side of the join. */ @@ -6366,7 +6374,7 @@ postgresGetForeignJoinPaths(PlannerInfo *root, */ static bool foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel, - Node *havingQual) + GroupPathExtraData *extra) { Query *query = root->parse; PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) grouped_rel->fdw_private; @@ -6375,6 +6383,7 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel, ListCell *lc; int i; List *tlist = NIL; + bool partial = extra->patype == PARTITIONWISE_AGGREGATE_PARTIAL; /* We currently don't support pushing Grouping Sets. */ if (query->groupingSets) @@ -6408,6 +6417,12 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel, * a node, as long as it's not at top level; then no match is possible. */ i = 0; + fpinfo->group_clause = query->groupClause; + if (partial) + { + fpinfo->group_clause = extra->groupClausePartial; + grouping_target = extra->partial_target; + } foreach(lc, grouping_target->exprs) { Expr *expr = (Expr *) lfirst(lc); @@ -6419,7 +6434,7 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel, * check the whole GROUP BY clause not just processed_groupClause, * because we will ship all of it, cf. appendGroupByClause. */ - if (sgref && get_sortgroupref_clause_noerr(sgref, query->groupClause)) + if (sgref && get_sortgroupref_clause_noerr(sgref, fpinfo->group_clause)) { TargetEntry *tle; @@ -6505,9 +6520,9 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel, * Classify the pushable and non-pushable HAVING clauses and save them in * remote_conds and local_conds of the grouped rel's fpinfo. */ - if (havingQual) + if (extra->havingQual) { - foreach(lc, (List *) havingQual) + foreach(lc, (List *) extra->havingQual) { Expr *expr = (Expr *) lfirst(lc); RestrictInfo *rinfo; @@ -6527,7 +6542,7 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel, grouped_rel->relids, NULL, NULL); - if (is_foreign_expr(root, grouped_rel, expr)) + if (is_foreign_expr(root, grouped_rel, expr) && !partial) fpinfo->remote_conds = lappend(fpinfo->remote_conds, rinfo); else fpinfo->local_conds = lappend(fpinfo->local_conds, rinfo); @@ -6546,6 +6561,12 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel, { RestrictInfo *rinfo = lfirst_node(RestrictInfo, lc); + if (partial) + { + rinfo = copyObject(rinfo); + lfirst(lc) = rinfo; + } + aggvars = list_concat(aggvars, pull_var_clause((Node *) rinfo->clause, PVC_INCLUDE_AGGREGATES)); @@ -6564,7 +6585,12 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel, */ if (IsA(expr, Aggref)) { - if (!is_foreign_expr(root, grouped_rel, expr)) + if (partial) + { + mark_partial_aggref((Aggref *) expr, AGGSPLIT_INITIAL_SERIAL); + continue; + } + else if (!is_foreign_expr(root, grouped_rel, expr)) return false; tlist = add_to_flat_tlist(tlist, list_make1(expr)); @@ -6621,6 +6647,7 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage, /* Ignore stages we don't support; and skip any duplicate calls. */ if ((stage != UPPERREL_GROUP_AGG && + stage != UPPERREL_PARTIAL_GROUP_AGG && stage != UPPERREL_ORDERED && stage != UPPERREL_FINAL) || output_rel->fdw_private) @@ -6637,6 +6664,10 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage, add_foreign_grouping_paths(root, input_rel, output_rel, (GroupPathExtraData *) extra); break; + case UPPERREL_PARTIAL_GROUP_AGG: + add_foreign_grouping_paths(root, input_rel, output_rel, + (GroupPathExtraData *) extra); + break; case UPPERREL_ORDERED: add_foreign_ordered_paths(root, input_rel, output_rel); break; @@ -6677,7 +6708,8 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel, return; Assert(extra->patype == PARTITIONWISE_AGGREGATE_NONE || - extra->patype == PARTITIONWISE_AGGREGATE_FULL); + extra->patype == PARTITIONWISE_AGGREGATE_FULL || + extra->patype == PARTITIONWISE_AGGREGATE_PARTIAL); /* save the input_rel as outerrel in fpinfo */ fpinfo->outerrel = input_rel; @@ -6697,7 +6729,7 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel, * Use HAVING qual from extra. In case of child partition, it will have * translated Vars. */ - if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual)) + if (!foreign_grouping_ok(root, grouped_rel, extra)) return; /* diff --git a/contrib/postgres_fdw/postgres_fdw.h b/contrib/postgres_fdw/postgres_fdw.h index 02c1152319..c60f3f5465 100644 --- a/contrib/postgres_fdw/postgres_fdw.h +++ b/contrib/postgres_fdw/postgres_fdw.h @@ -86,6 +86,15 @@ typedef struct PgFdwRelationInfo ForeignServer *server; UserMapping *user; /* only set in use_remote_estimate mode */ + /* for partial aggregate pushdown */ + bool check_partial_aggregate_support; + + /* + * if remoteversion is zero, then that means the remote server version is + * unacquired. + */ + int remoteversion; + int fetch_size; /* fetch size for this remote table */ /* @@ -110,6 +119,7 @@ typedef struct PgFdwRelationInfo /* Grouping information */ List *grouped_tlist; + List *group_clause; /* Subquery information */ bool make_outerrel_subquery; /* do we deparse outerrel as a diff --git a/contrib/postgres_fdw/sql/postgres_fdw.sql b/contrib/postgres_fdw/sql/postgres_fdw.sql index b54903ad8f..1824cb67fe 100644 --- a/contrib/postgres_fdw/sql/postgres_fdw.sql +++ b/contrib/postgres_fdw/sql/postgres_fdw.sql @@ -2963,16 +2963,31 @@ RESET enable_partitionwise_join; -- =================================================================== -- test partitionwise aggregates -- =================================================================== - -CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a); +ALTER SERVER loopback OPTIONS (ADD fdw_tuple_cost '0.1'); + +CREATE TYPE mood AS enum ('sad', 'ok', 'happy'); +ALTER EXTENSION postgres_fdw ADD TYPE mood; + +CREATE TABLE pagg_tab (a int, b int, c text, c_serial int4, + c_int4array _int4, c_interval interval, + c_money money, c_1c text, c_1b bytea, + c_bit bit(2), c_1or3int2 int2, + c_1or3int4 int4, c_1or3int8 int8, + c_bool bool, c_enum mood, c_pg_lsn pg_lsn, + c_tid tid, c_int4range int4range, + c_int4multirange int4multirange, + c_time time, c_timetz timetz, + c_timestamp timestamp, c_timestamptz timestamptz, + c_xid8 xid8) + PARTITION BY RANGE(a); CREATE TABLE pagg_tab_p1 (LIKE pagg_tab); CREATE TABLE pagg_tab_p2 (LIKE pagg_tab); CREATE TABLE pagg_tab_p3 (LIKE pagg_tab); -INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10; -INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10; -INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20; +INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i, array[(i % 2), 0], ((i % 2) || ' seconds')::interval, (i % 2)::money, ((i - 1) % 10)::text, (((i - 1) % 10)::text)::bytea, case when (i % 2) = 0 then B'01' else B'11' end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then true else false end, (case when (i % 2) = 0 then 'sad' else 'happy' end)::mood, ('0/' || i)::pg_lsn, ('(0,' || i || ')')::tid, int4range(0, i), int4multirange(int4range(0, i), int4range(100, 100 + i)), '00:00:00'::time + (i || ' seconds')::interval, '00:00:00'::timetz + (i || ' seconds')::interval, '2000-01-01'::timestamp + (i || ' seconds')::interval, '2000-01-01'::timestamptz + (i || ' seconds')::interval, ((i % 10)::text)::xid8 FROM generate_series(1, 3000) i WHERE (i % 30) < 10; +INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i, array[(i % 2), 0], ((i % 2) || ' seconds')::interval, (i % 2)::money, ((i - 1) % 10)::text, (((i - 1) % 10)::text)::bytea, case when (i % 2) = 0 then B'01' else B'11' end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then true else false end, (case when (i % 2) = 0 then 'sad' else 'happy' end)::mood, ('0/' || i)::pg_lsn, ('(0,' || i || ')')::tid, int4range(0, i), int4multirange(int4range(0, i), int4range(100, 100 + i)), '00:00:00'::time + (i || ' seconds')::interval, '00:00:00'::timetz + (i || ' seconds')::interval, '2000-01-01'::timestamp + (i || ' seconds')::interval, '2000-01-01'::timestamptz + (i || ' seconds')::interval, ((i % 10)::text)::xid8 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10; +INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i, array[(i % 2), 0], ((i % 2) || ' seconds')::interval, (i % 2)::money, ((i - 1) % 10)::text, (((i - 1) % 10)::text)::bytea, case when (i % 2) = 0 then B'01' else B'11' end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then true else false end, (case when (i % 2) = 0 then 'sad' else 'happy' end)::mood, ('0/' || i)::pg_lsn, ('(0,' || i || ')')::tid, int4range(0, i), int4multirange(int4range(0, i), int4range(100, 100 + i)), '00:00:00'::time + (i || ' seconds')::interval, '00:00:00'::timetz + (i || ' seconds')::interval, '2000-01-01'::timestamp + (i || ' seconds')::interval, '2000-01-01'::timestamptz + (i || ' seconds')::interval, ((i % 10)::text)::xid8 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20; -- Create foreign partitions CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1'); @@ -2980,6 +2995,9 @@ CREATE FOREIGN TABLE fpagg_tab_p2 PARTITION OF pagg_tab FOR VALUES FROM (10) TO CREATE FOREIGN TABLE fpagg_tab_p3 PARTITION OF pagg_tab FOR VALUES FROM (20) TO (30) SERVER loopback OPTIONS (table_name 'pagg_tab_p3'); ANALYZE pagg_tab; +ANALYZE pagg_tab_p1; +ANALYZE pagg_tab_p2; +ANALYZE pagg_tab_p3; ANALYZE fpagg_tab_p1; ANALYZE fpagg_tab_p2; ANALYZE fpagg_tab_p3; @@ -3002,10 +3020,289 @@ EXPLAIN (VERBOSE, COSTS OFF) SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1; SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1; --- When GROUP BY clause does not match with PARTITION KEY. -EXPLAIN (COSTS OFF) +-- Partial aggregates are safe to push down having clause when there are partial aggregates +EXPLAIN (VERBOSE, COSTS OFF) +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1; SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1; +-- Partial aggregates are safe to push down without having clause +EXPLAIN (VERBOSE, COSTS OFF) +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; + +-- Partial aggregates are safe to push down even if we need both variable and variable-based expression +EXPLAIN (VERBOSE, COSTS OFF) +SELECT (b/2)::numeric, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b/2 ORDER BY 1; +SELECT (b/2)::numeric, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b/2 ORDER BY 1; + +-- Partial aggregates are safe to push down for all built-in aggregates +EXPLAIN (VERBOSE, COSTS OFF) +SELECT /* aggregate function <> aggpartialfunc */ + array_agg(c_int4array), array_agg(b), + avg(b::int2), avg(b::int4), avg(b::int8), avg(c_interval), avg(b::float4), avg(b::float8), avg(b::numeric), + corr(b::float8, (b * b)::float8), + covar_pop(b::float8, (b * b)::float8), + covar_samp(b::float8, (b * b)::float8), + regr_avgx((2 * b)::float8, b::float8), + regr_avgy((2 * b)::float8, b::float8), + regr_intercept((2 * b + 3)::float8, b::float8), + regr_r2((b * b)::float8, b::float8), + regr_slope((2 * b + 3)::float8, b::float8), + regr_sxx((2 * b + 3)::float8, b::float8), + regr_sxy((b * b)::float8, b::float8), + regr_syy((2 * b + 3)::float8, b::float8), + stddev(b::float4), stddev(b::float8), stddev(b::int2), stddev(b::int4), stddev(b::int8), stddev(b::numeric), + stddev_pop(b::float4), stddev_pop(b::float8), stddev_pop(b::int2), stddev_pop(b::int4), stddev_pop(b::int8), stddev_pop(b::numeric), + stddev_samp(b::float4), stddev_samp(b::float8), stddev_samp(b::int2), stddev_samp(b::int4), stddev_samp(b::int8), stddev_samp(b::numeric), + string_agg(c_1c, ','), string_agg(c_1b, ','), + sum(b::int8), sum(b::numeric), + variance(b::float4), variance(b::float8), variance(b::int2), variance(b::int4), variance(b::int8), variance(b::numeric), + var_pop(b::float4), var_pop(b::float8), var_pop(b::int2), var_pop(b::int4), var_pop(b::int8), var_pop(b::numeric), + var_samp(b::float4), var_samp(b::float8), var_samp(b::int2), var_samp(b::int4), var_samp(b::int8), var_samp(b::numeric), + /* aggregate function = aggpartialfunc */ + any_value(b * 0), + bit_and(c_bit), bit_and(c_1or3int2), bit_and(c_1or3int4), bit_and(c_1or3int8), + bit_or(c_bit), bit_or(c_1or3int2), bit_or(c_1or3int4), bit_or(c_1or3int8), + bit_xor(c_bit), bit_xor(c_1or3int2), bit_xor(c_1or3int4), bit_xor(c_1or3int8), + bool_and(c_bool), + bool_or(c_bool), + count(b), count(*), + every(c_bool), + max(c_int4array), max(c_enum), max(c_1c::char(1)), max('2000-01-01'::date + b), max('0.0.0.0'::inet + b), max(b::float4), max(b::float8), max(b::int2), max(b::int4), max(b::int8), max(c_interval), max(c_money), max(b::numeric), max(b::oid), max(c_pg_lsn), max(c_tid), max(c_1c), max(c_time), max(c_timetz), max(c_timestamp), max(c_timestamptz), max(c_xid8), + min(c_int4array), min(c_enum), min(c_1c::char(1)), min('2000-01-01'::date + b), min('0.0.0.0'::inet + b), min(b::float4), min(b::float8), min(b::int2), min(b::int4), min(b::int8), min(c_interval), min(c_money), min(b::numeric), min(b::oid), min(c_pg_lsn), min(c_tid), min(c_1c), min(c_time), min(c_timetz), min(c_timestamp), min(c_timestamptz), min(c_xid8), + range_intersect_agg(c_int4range), range_intersect_agg(c_int4multirange), + regr_count((2 * b + 3)::float8, b::float8), + sum(b::float4), sum(b::float8), sum(b::int2), sum(b::int4), sum(c_interval), sum(c_money) + FROM pagg_tab WHERE c_serial between 1 and 30; + +SELECT /* aggregate function <> aggpartialfunc */ + array_agg(c_int4array), array_agg(b), + avg(b::int2), avg(b::int4), avg(b::int8), avg(c_interval), + avg(b::float4), avg(b::float8), + corr(b::float8, (b * b)::float8), + covar_pop(b::float8, (b * b)::float8), + covar_samp(b::float8, (b * b)::float8), + regr_avgx((2 * b)::float8, b::float8), + regr_avgy((2 * b)::float8, b::float8), + regr_intercept((2 * b + 3)::float8, b::float8), + regr_r2((b * b)::float8, b::float8), + regr_slope((2 * b + 3)::float8, b::float8), + regr_sxx((2 * b + 3)::float8, b::float8), + regr_sxy((b * b)::float8, b::float8), + regr_syy((2 * b + 3)::float8, b::float8), + stddev(b::float4), stddev(b::float8), stddev(b::int2), stddev(b::int4), stddev(b::int8), stddev(b::numeric), + stddev_pop(b::float4), stddev_pop(b::float8), stddev_pop(b::int2), stddev_pop(b::int4), stddev_pop(b::int8), stddev_pop(b::numeric), + stddev_samp(b::float4), stddev_samp(b::float8), stddev_samp(b::int2), stddev_samp(b::int4), stddev_samp(b::int8), stddev_samp(b::numeric), + string_agg(c_1c, ','), string_agg(c_1b, ','), + sum(b::int8), sum(b::numeric), + variance(b::float4), variance(b::float8), variance(b::int2), variance(b::int4), variance(b::int8), variance(b::numeric), + var_pop(b::float4), var_pop(b::float8), var_pop(b::int2), var_pop(b::int4), var_pop(b::int8), var_pop(b::numeric), + var_samp(b::float4), var_samp(b::float8), var_samp(b::int2), var_samp(b::int4), var_samp(b::int8), var_samp(b::numeric), + /* aggregate function = aggpartialfunc */ + any_value(b * 0), + bit_and(c_bit), bit_and(c_1or3int2), bit_and(c_1or3int4), bit_and(c_1or3int8), + bit_or(c_bit), bit_or(c_1or3int2), bit_or(c_1or3int4), bit_or(c_1or3int8), + bit_xor(c_bit), bit_xor(c_1or3int2), bit_xor(c_1or3int4), bit_xor(c_1or3int8), + bool_and(c_bool), + bool_or(c_bool), + count(b), count(*), + every(c_bool), + max(c_int4array), max(c_enum), max(c_1c::char(1)), max('2000-01-01'::date + b), max('0.0.0.0'::inet + b), max(b::float4), max(b::float8), max(b::int2), max(b::int4), max(b::int8), max(c_interval), max(c_money), max(b::numeric), max(b::oid), max(c_pg_lsn), max(c_tid), max(c_1c), max(c_time), max(c_timetz), max(c_timestamp), max(c_timestamptz), max(c_xid8), + min(c_int4array), min(c_enum), min(c_1c::char(1)), min('2000-01-01'::date + b), min('0.0.0.0'::inet + b), min(b::float4), min(b::float8), min(b::int2), min(b::int4), min(b::int8), min(c_interval), min(c_money), min(b::numeric), min(b::oid), min(c_pg_lsn), min(c_tid), min(c_1c), min(c_time), min(c_timetz), min(c_timestamp), min(c_timestamptz), min(c_xid8), + range_intersect_agg(c_int4range), range_intersect_agg(c_int4multirange), + regr_count((2 * b + 3)::float8, b::float8), + sum(b::float4), sum(b::float8), sum(b::int2), sum(b::int4), sum(c_interval), sum(c_money) + FROM pagg_tab WHERE c_serial between 1 and 30; + +-- Tests for backward compatibility +ALTER SERVER loopback OPTIONS (ADD check_partial_aggregate_support 'false'); +EXPLAIN (VERBOSE, COSTS OFF) +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; + +ALTER SERVER loopback OPTIONS (SET check_partial_aggregate_support 'true'); +EXPLAIN (VERBOSE, COSTS OFF) +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; +SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; + +ALTER SERVER loopback OPTIONS (SET fdw_tuple_cost '1.0'); +SET enable_partitionwise_join=on; +EXPLAIN (VERBOSE, COSTS OFF) +SELECT avg(t1.b), avg(t1.b::int8) FROM pagg_tab t1 JOIN pagg_tab t2 USING(a); +SELECT avg(t1.b), avg(t1.b::int8) FROM pagg_tab t1 JOIN pagg_tab t2 USING(a); +RESET enable_partitionwise_join; +ALTER SERVER loopback OPTIONS (SET fdw_tuple_cost '0.1'); + +ALTER SERVER loopback OPTIONS (DROP check_partial_aggregate_support); + +-- List of aggregate functions such that the aggregate function supports partial aggregate +-- and aggpartialfn field of the aggregate function is invalid +SELECT aggfnoid::text + FROM pg_aggregate a JOIN pg_type b ON a.aggtranstype = b.oid + WHERE (aggcombinefn::text <> '-') AND (aggpartialfn::text = '-') + ORDER BY 1; + +-- List of aggregate functions which have incorrect aggpartialfunc +SELECT a.aggfnoid + FROM pg_aggregate a JOIN pg_type b ON a.aggtranstype = b.oid + JOIN pg_aggregate c on a.aggpartialfn = c.aggfnoid + WHERE (a.aggcombinefn::text <> '-') + AND (a.aggpartialfn::text <> '-') + AND (a.aggtransfn <> c.aggtransfn + OR (c.aggcombinefn::text <> '-' + AND a.aggcombinefn <> c.aggcombinefn) + OR (a.agginitval IS NOT NULL + AND c.agginitval IS NOT NULL + AND a.agginitval <> c.agginitval) + OR (b.typname = 'internal' + AND c.aggserialfn::text <> '-' + AND a.aggserialfn <> c.aggserialfn) + OR (b.typname = 'internal' + AND c.aggdeserialfn::text <> '-' + AND a.aggdeserialfn <> c.aggdeserialfn) + OR (b.typname = 'internal' + AND a.aggserialfn <> c.aggfinalfn) + OR (b.typname <> 'internal' + AND c.aggfinalfn::text <> '-')) + ORDER BY 1; + +-- It's unsafe to push down partial aggregates which contain distinct clause +EXPLAIN (VERBOSE, COSTS OFF) +SELECT max(a), count(distinct b) FROM pagg_tab; +SELECT max(a), count(distinct b) FROM pagg_tab; + +-- It's unsafe to push down partial aggregates which contain order by clause +EXPLAIN (VERBOSE, COSTS OFF) SELECT array_agg(b order by b) FROM pagg_tab WHERE c_serial between 1 and 30; +SELECT array_agg(b order by b) FROM pagg_tab WHERE c_serial between 1 and 30; + +-- Create user-defined aggregates whose stype is internal +CREATE SCHEMA postgres_fdw_test; +CREATE AGGREGATE postgres_fdw_test.postgres_fdw_sum_p_int8(int8) ( + SFUNC = int8_avg_accum, + STYPE = internal, + COMBINEFUNC = int8_avg_combine, + FINALFUNC = int8_avg_serialize, + SERIALFUNC = int8_avg_serialize, + DESERIALFUNC = int8_avg_deserialize, + AGGPARTIALFUNC = postgres_fdw_test.postgres_fdw_sum_p_int8 +); + +CREATE AGGREGATE postgres_fdw_sum(int8) ( + SFUNC = int8_avg_accum, + STYPE = internal, + COMBINEFUNC = int8_avg_combine, + FINALFUNC = numeric_poly_sum, + SERIALFUNC = int8_avg_serialize, + DESERIALFUNC = int8_avg_deserialize, + AGGPARTIALFUNC = postgres_fdw_test.postgres_fdw_sum_p_int8 +); + +-- Create user-defined aggregates whose stype is not internal +CREATE AGGREGATE postgres_fdw_avg_p_int4(int4) ( + SFUNC = int4_avg_accum, + STYPE = _int8, + COMBINEFUNC = int4_avg_combine, + INITCOND = '{0,0}', + AGGPARTIALFUNC = postgres_fdw_avg_p_int4 +); + +CREATE AGGREGATE postgres_fdw_avg(int4) ( + SFUNC = int4_avg_accum, + STYPE = _int8, + COMBINEFUNC = int4_avg_combine, + FINALFUNC = int8_avg, + INITCOND = '{0,0}', + AGGPARTIALFUNC = postgres_fdw_avg_p_int4 +); + +CREATE AGGREGATE postgres_fdw_sum_noaggpartialfunc(int8) ( + SFUNC = int8_avg_accum, + STYPE = internal, + COMBINEFUNC = int8_avg_combine, + FINALFUNC = numeric_poly_sum, + SERIALFUNC = int8_avg_serialize, + DESERIALFUNC = int8_avg_deserialize +); + +-- It's safe to push down partial aggregate for user-defined aggregate +-- when it belongs to an extension that's listed in extensions set +-- and when its aggpartialfunc belongs to an extension that's listed +-- in extensions set. +ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_sum(int8); +ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_test.postgres_fdw_sum_p_int8(int8); + +ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_avg(int4); +ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_avg_p_int4(int4); + +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_sum(b::int8) FROM pagg_tab; +SELECT postgres_fdw_sum(b::int8) FROM pagg_tab; + +-- It's unsafe to push down partial aggregate for user-defined aggregate +-- when it has no aggpartialfunc. +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_sum_noaggpartialfunc(b::int8) FROM pagg_tab; +SELECT postgres_fdw_sum_noaggpartialfunc(b::int8) FROM pagg_tab; + +ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_test.postgres_fdw_sum_p_int8(int8); + +-- Reconnect for flushing the shippability cache +\c - +SET enable_partitionwise_aggregate TO true; + +-- It's unsafe to push down partial aggregate for user-defined aggregate +-- when itself is aggpartialfunc and when it belongs to an extension +-- that's listed in extensions set and when stype is internal. +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_test.postgres_fdw_sum_p_int8(b::int8) FROM pagg_tab; +SELECT postgres_fdw_test.postgres_fdw_sum_p_int8(b::int8) FROM pagg_tab; + +ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_sum(int8); +ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_test.postgres_fdw_sum_p_int8(int8); + +-- Reconnect for flushing the shippability cache +\c - +SET enable_partitionwise_aggregate TO true; + +-- It's unsafe to push down partial aggregate for user-defined aggregate +-- when it doesn't belong to an extension that's listed in extensions set +-- and when its aggpartialfunc belongs to an extension that's listed +-- in extensions set. +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_sum(b::int8) FROM pagg_tab; +SELECT postgres_fdw_sum(b::int8) FROM pagg_tab; + +-- It's safe to push down partial aggregate for user-defined aggregate +-- whose stype is not internal +-- when it belongs to an extension that's listed in extensions set +-- and when its aggpartialfunc belongs to an extension that's listed +-- in extensions set. +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_avg(b) FROM pagg_tab; +SELECT postgres_fdw_avg(b) FROM pagg_tab; + +-- It's safe to push down partial aggregate for user-defined aggregate +-- when itself is aggpartialfunc and when it belongs to an extension +-- that's listed in extensions set and when stype is not internal. +EXPLAIN (VERBOSE, COSTS OFF) +SELECT postgres_fdw_avg_p_int4(b) FROM pagg_tab; +SELECT postgres_fdw_avg_p_int4(b) FROM pagg_tab; + +ALTER EXTENSION postgres_fdw DROP TYPE mood; + +ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_test.postgres_fdw_sum_p_int8(int8); + +ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_avg_p_int4(int4); +ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_avg(int4); + +DROP AGGREGATE postgres_fdw_sum(int8); +DROP AGGREGATE postgres_fdw_test.postgres_fdw_sum_p_int8(int8); +DROP SCHEMA postgres_fdw_test; +DROP AGGREGATE postgres_fdw_avg(int4); +DROP AGGREGATE postgres_fdw_avg_p_int4(int4); +DROP AGGREGATE postgres_fdw_sum_noaggpartialfunc(int8); +ALTER SERVER loopback OPTIONS (DROP fdw_tuple_cost); + -- =================================================================== -- access rights and superuser -- =================================================================== diff --git a/doc/src/sgml/catalogs.sgml b/doc/src/sgml/catalogs.sgml index ed32ca0349..4f5ff32e96 100644 --- a/doc/src/sgml/catalogs.sgml +++ b/doc/src/sgml/catalogs.sgml @@ -644,6 +644,19 @@ value starts out null. </para></entry> </row> + + <row> + <entry role="catalog_table_entry"><para role="column_definition"> + <structfield>aggpartialfn</structfield> <type>regproc</type> + (references <link linkend="catalog-pg-proc"><structname>pg_proc</structname></link>.<structfield>oid</structfield>) + </para> + <para> + Partial aggregate function (zero if none). + See + <xref linkend="partial-aggregate-pushdown"/> for the + definition of partial aggregate function. + </para></entry> + </row> </tbody> </tgroup> </table> diff --git a/doc/src/sgml/postgres-fdw.sgml b/doc/src/sgml/postgres-fdw.sgml index 5062d712e7..676abc6fa5 100644 --- a/doc/src/sgml/postgres-fdw.sgml +++ b/doc/src/sgml/postgres-fdw.sgml @@ -451,6 +451,37 @@ OPTIONS (ADD password_required 'false'); </sect3> + <sect3 id="postgres-fdw-options-partial-aggregate-pushdown"> + <title>Partial Aggregate Pushdown Options</title> + + <para> + By default, <filename>postgres_fdw</filename> assumes that for each built-in aggregate + function, the partial aggregate function is defined on the remote server. See + <xref linkend="partial-aggregate-pushdown"/> for the definition of the partial aggregate + function. This may be overridden using the following option: + </para> + + <variablelist> + + <varlistentry> + <term><literal>check_partial_aggregate_support</literal> (<type>boolean</type>)</term> + <listitem> + <para> + When this option is true, <filename>postgres_fdw</filename> connects to the remote server + and gets its remote server version during query planning. + If the remote server version is older than the local server version, + <filename>postgres_fdw</filename> + assumes that for each aggregate function, the partial aggregate function is not defined + on the remote server unless the partial aggregate function and the aggregate + function match. + The default is <literal>false</literal>. + </para> + </listitem> + </varlistentry> + + </variablelist> + </sect3> + <sect3 id="postgres-fdw-options-asynchronous-execution"> <title>Asynchronous Execution Options</title> @@ -929,13 +960,15 @@ postgres=# SELECT postgres_fdw_disconnect_all(); <para> <filename>postgres_fdw</filename> attempts to optimize remote queries to reduce the amount of data transferred from foreign servers. This is done by - sending query <literal>WHERE</literal> clauses to the remote server for - execution, and by not retrieving table columns that are not needed for - the current query. To reduce the risk of misexecution of queries, - <literal>WHERE</literal> clauses are not sent to the remote server unless they use - only data types, operators, and functions that are built-in or belong to an - extension that's listed in the foreign server's <literal>extensions</literal> - option. Operators and functions in such clauses must + sending query <literal>WHERE</literal> clauses and aggregate expressions + to the remote server for execution, and by not retrieving table columns that + are not needed for the current query. + To reduce the risk of misexecution of queries, + <literal>WHERE</literal> clauses and aggregate expressions are not sent to + the remote server unless they use only data types, operators, and functions + that are built-in or belong to an extension that's listed in the foreign + server's <literal>extensions</literal> option. + Operators and functions in such clauses must be <literal>IMMUTABLE</literal> as well. For an <command>UPDATE</command> or <command>DELETE</command> query, <filename>postgres_fdw</filename> attempts to optimize the query execution by @@ -966,6 +999,68 @@ postgres=# SELECT postgres_fdw_disconnect_all(); </para> </sect2> + <sect2 id="partial-aggregate-pushdown"> + <title>Partial aggregate pushdown</title> + <para> + Partial aggregate pushdown is an optimization for a query that contains aggregate + expressions for a partitioned table across one or more remote servers. If the multiple + conditions are true for a remote server, an aggregate expression is sent to that + remote server and the aggregate is a partial aggregate on the local server. + The conditions under which this sending is active are as follows. + <itemizedlist spacing="compact"> + <listitem> + <para> + An aggregate function, called the partial aggregate function for partial aggregate + that corresponding to the aggregate expression, is defined on the local server and + the remote server. + </para> + </listitem> + <listitem> + <para> + The conditions in <xref linkend="postgres-fdw-remote-query-optimization"/> are true. + </para> + </listitem> + <listitem> + <para> + The query doesn't contain a <literal>HAVING</literal> clause. + </para> + </listitem> + <listitem> + <para> + The aggregate expressions in the query don't contain + a <literal>DISTINCT</literal> or <literal>ORDER BY</literal> clause. + </para> + </listitem> + </itemizedlist> + When this sending is active for aggregate expressions, any aggregate + function is replaced by the partial aggregate function in the remote query. + If the data type of the state value is not internal, the partial aggregate + function has no final function. + Otherwise, the partial aggregate function must have the same final function + as the aggregate function's serialization function. + The partial aggregate function and the aggregate function have + the same transition function and data type of the state value. + If the partial aggregate function has any of the following members: + initial state, combine function, serialization function, or deserialization + function, the members must match with the corresponding members in the + aggregate function. + For all built-in aggregate functions that support partial aggregation, the partial + aggregate function that corresponds to this function is defined as a built-in + aggregate function. + By default, <filename>postgres_fdw</filename> assumes that for each built-in + aggregate function, the partial aggregate function is defined on the local server + and the remote server. + If <literal>check_partial_aggregate_support</literal> is true and the remote server version is + older than the local server version, <filename>postgres_fdw</filename> does not assume + that the partial aggregate function is on the remote server unless + the partial aggregate function and the aggregate function match. + <filename>postgres_fdw</filename> assumes that a user-defined aggregate + function has the partial aggregate function on the remote server + only if both belong to an extension that is listed in the foreign server's + extensions option. + </para> + </sect2> + <sect2 id="postgres-fdw-remote-query-execution-environment"> <title>Remote Query Execution Environment</title> @@ -1031,14 +1126,27 @@ postgres=# SELECT postgres_fdw_disconnect_all(); back to 8.1. A limitation however is that <filename>postgres_fdw</filename> generally assumes that immutable built-in functions and operators are safe to send to the remote server for execution, if they appear in a - <literal>WHERE</literal> clause for a foreign table. Thus, a built-in - function that was added since the remote server's release might be sent - to it for execution, resulting in <quote>function does not exist</quote> or - a similar error. This type of failure can be worked around by + <literal>WHERE</literal> clause or aggregate expressions for a foreign table. + Thus, a built-in function that was added since the remote server's release + might be sent to it for execution, resulting in <quote>function does not + exist</quote> or a similar error. This type of failure can be worked around by rewriting the query, for example by embedding the foreign table reference in a sub-<literal>SELECT</literal> with <literal>OFFSET 0</literal> as an optimization fence, and placing the problematic function or operator outside the sub-<literal>SELECT</literal>. + A similar problem occurs when using the partial aggregate pushdown feature. + When using the partial aggregate pushdown feature, + the partial aggregate function of the aggregate function is sent to + the remote server if the conditions in + <xref linkend="partial-aggregate-pushdown"/> are met. + A limitation is that by default, <filename>postgres_fdw</filename> assumes that + for each built-in aggregate function, the partial aggregate function is defined + on the remote server. + This limitation might cause <quote>function does not exist</quote> + or a similar error if the remote server version is older than the local server version. + This problem can be worked around by setting + <literal>check_partial_aggregate_support</literal> to <literal>true</literal> + or similar query rewriting as above. </para> </sect2> diff --git a/doc/src/sgml/ref/create_aggregate.sgml b/doc/src/sgml/ref/create_aggregate.sgml index 222e0aa5c9..2875734dea 100644 --- a/doc/src/sgml/ref/create_aggregate.sgml +++ b/doc/src/sgml/ref/create_aggregate.sgml @@ -42,6 +42,7 @@ CREATE [ OR REPLACE ] AGGREGATE <replaceable class="parameter">name</replaceable [ , MINITCOND = <replaceable class="parameter">minitial_condition</replaceable> ] [ , SORTOP = <replaceable class="parameter">sort_operator</replaceable> ] [ , PARALLEL = { SAFE | RESTRICTED | UNSAFE } ] + [ , AGGPARTIALFUNC = <replaceable class="parameter">aggpartialfunc</replaceable> ] ) CREATE [ OR REPLACE ] AGGREGATE <replaceable class="parameter">name</replaceable> ( [ [ <replaceable class="parameter">argmode</replaceable> ] [ <replaceable class="parameter">argname</replaceable> ] <replaceable class="parameter">arg_data_type</replaceable> [ , ... ] ] @@ -80,6 +81,7 @@ CREATE [ OR REPLACE ] AGGREGATE <replaceable class="parameter">name</replaceable [ , MFINALFUNC_MODIFY = { READ_ONLY | SHAREABLE | READ_WRITE } ] [ , MINITCOND = <replaceable class="parameter">minitial_condition</replaceable> ] [ , SORTOP = <replaceable class="parameter">sort_operator</replaceable> ] + [ , AGGPARTIALFUNC = <replaceable class="parameter">aggpartialfunc</replaceable> ] ) </synopsis> </refsynopsisdiv> @@ -246,6 +248,18 @@ CREATE [ OR REPLACE ] AGGREGATE <replaceable class="parameter">name</replaceable marked <literal>PARALLEL SAFE</literal> to enable parallel aggregation. </para> + <para> + <filename>postgres_fdw</filename> can optimize a query containing aggregate + expressions for a partitioned table across one or more remote servers. + This optimization sends an aggregate expression to a remote server and the + aggregate is a partial aggregate on the local server. This optimization + requires an aggregate function called a <firstterm>partial aggregate + function</firstterm>. See + <xref linkend="partial-aggregate-pushdown"/> for the definition of + partial aggregate function. <literal>AGGPARTIALFUNC</literal> is a parameter + of the partial aggregate function. + </para> + <para> Aggregates that behave like <function>MIN</function> or <function>MAX</function> can sometimes be optimized by looking into an index instead of scanning every @@ -653,6 +667,34 @@ SELECT col FROM tab ORDER BY col USING sortop LIMIT 1; </para> </listitem> </varlistentry> + + <varlistentry> + <term><replaceable class="parameter">aggpartialfunc</replaceable></term> + <listitem> + <para> + The name of the partial aggregate function. + If the <replaceable class="parameter">name</replaceable> and the + <replaceable class="parameter">aggpartialfunc</replaceable> are the same, + then the aggregate function must be the partial aggregate function. + If the <replaceable class="parameter">aggpartialfunc</replaceable> + is specified, the <replaceable class="parameter">combinefunc</replaceable> + must be specified. + If the <replaceable class="parameter">state_data_type</replaceable> is not + <type>internal</type>, then the partial aggregate function must not have + the <replaceable class="parameter">ffunc</replaceable>. + If the <replaceable class="parameter">state_data_type</replaceable> is + <type>internal</type>, the partial aggregate function must have the same + final function as the <replaceable class="parameter">serialfunc</replaceable>. + The partial aggregate function and the aggregate function have + the same <replaceable class="parameter">sfunc</replaceable> and + <replaceable class="parameter">state_data_type</replaceable>. + If the partial aggregate function has any of the following members: + initial state, combine function, serialization function, or deserialization + function, the members must match with the corresponding members in the + aggregate function. + </para> + </listitem> + </varlistentry> </variablelist> <para> diff --git a/doc/src/sgml/xaggr.sgml b/doc/src/sgml/xaggr.sgml index bdad8d3dc2..607c859974 100644 --- a/doc/src/sgml/xaggr.sgml +++ b/doc/src/sgml/xaggr.sgml @@ -536,9 +536,6 @@ SELECT percentile_disc(0.5) WITHIN GROUP (ORDER BY income) FROM households; parallel aggregation by having different worker processes scan different portions of a table. Each worker produces a partial state value, and at the end those state values are combined to produce a final state value. - (In the future this mode might also be used for purposes such as combining - aggregations over local and remote tables; but that is not implemented - yet.) </para> <para> @@ -612,6 +609,18 @@ SELECT percentile_disc(0.5) WITHIN GROUP (ORDER BY income) FROM households; parallel-safety markings on its support functions are not consulted. </para> + <para> + <filename>postgres_fdw</filename> can optimize a query containing + aggregate expressions for a partitioned table across one or more + remote servers. This optimization sends an aggregate expression to + a remote server and the aggregate is a partial aggregate on the + local server. This optimization requires an aggregate function + called the <firstterm>partial aggregate function</firstterm>. + See + <xref linkend="partial-aggregate-pushdown"/> for the + definition of the partial aggregate function. + </para> + </sect2> <sect2 id="xaggr-support-functions"> diff --git a/src/backend/catalog/pg_aggregate.c b/src/backend/catalog/pg_aggregate.c index ebc4454743..aaf1e62762 100644 --- a/src/backend/catalog/pg_aggregate.c +++ b/src/backend/catalog/pg_aggregate.c @@ -63,6 +63,7 @@ AggregateCreate(const char *aggName, List *aggmtransfnName, List *aggminvtransfnName, List *aggmfinalfnName, + List *aggpartialfnName, bool finalfnExtraArgs, bool mfinalfnExtraArgs, char finalfnModify, @@ -91,6 +92,8 @@ AggregateCreate(const char *aggName, Oid mtransfn = InvalidOid; /* can be omitted */ Oid minvtransfn = InvalidOid; /* can be omitted */ Oid mfinalfn = InvalidOid; /* can be omitted */ + bool isaggpartialfn = false; + Oid aggpartialfn = InvalidOid; /* can be omitted */ Oid sortop = InvalidOid; /* can be omitted */ Oid *aggArgTypes = parameterTypes->values; bool mtransIsStrict = false; @@ -569,6 +572,94 @@ AggregateCreate(const char *aggName, format_type_be(finaltype)))); } + /* + * Validate the aggpartialfunc, if present. + */ + if (aggpartialfnName) + { + char *aggpartialName; + Oid aggpartialNamespace; + + if (!aggcombinefnName) + elog(ERROR, "aggcombinefnName must be supplied if aggpartialfnName is supplied"); + + /* Convert list of names to a name and namespace */ + aggpartialNamespace = QualifiedNameGetCreationNamespace(aggpartialfnName, + &aggpartialName); + + if ((aggNamespace == aggpartialNamespace) + && (strcmp(aggName, aggpartialName) == 0)) + { + if (((aggTransType != INTERNALOID) && (finalfn != InvalidOid)) + || ((aggTransType == INTERNALOID) && (finalfn != serialfn))) + elog(ERROR, "%s is not its own aggpartialfunc", aggName); + isaggpartialfn = true; + } + else + { + HeapTuple aggtup; + Form_pg_aggregate aggpartialform; + Datum textInitVal; + char *strInitVal; + bool initValueIsNull; + + aggpartialfn = LookupFuncName(aggpartialfnName, numArgs, aggArgTypes, false); + + /* Check aggregate creator has permission to call the function */ + aclresult = object_aclcheck(ProcedureRelationId, aggpartialfn, + GetUserId(), ACL_EXECUTE); + if (aclresult != ACLCHECK_OK) + aclcheck_error(aclresult, OBJECT_FUNCTION, get_func_name(aggpartialfn)); + + rettype = get_func_rettype(aggpartialfn); + + aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(aggpartialfn)); + if (!HeapTupleIsValid(aggtup)) + elog(ERROR, "cache lookup failed for aggpartialfunc %u", aggpartialfn); + + aggpartialform = (Form_pg_aggregate) GETSTRUCT(aggtup); + + if ((aggpartialform->aggcombinefn != InvalidOid) && + (aggpartialform->aggcombinefn != combinefn)) + elog(ERROR, "%s in aggregate and %s in its aggpartialfunc must match", + "combinefunc", "combinefunc"); + + if (aggpartialform->aggtransfn != transfn) + elog(ERROR, "%s in aggregate and its aggpartialfunc must match", "sfunc"); + + textInitVal = SysCacheGetAttr(AGGFNOID, aggtup, + Anum_pg_aggregate_agginitval, &initValueIsNull); + + if (!initValueIsNull && (agginitval != NULL)) + { + strInitVal = TextDatumGetCString(textInitVal); + if (strcmp(strInitVal, agginitval) != 0) + elog(ERROR, "%s in aggregate and %s in its aggpartialfunc must match", + "initcond", "initcond"); + } + + if (aggTransType == INTERNALOID) + { + if (aggpartialform->aggcombinefn != InvalidOid) + { + if (aggpartialform->aggserialfn != serialfn) + elog(ERROR, "%s in aggregate and %s in its aggpartialfunc must match", + "serialfunc", "serialfunc"); + + if (aggpartialform->aggdeserialfn != deserialfn) + elog(ERROR, "%s in aggregate and %s in its aggpartialfunc must match", + "deserialfunc", "deserialfunc"); + } + if (aggpartialform->aggfinalfn != serialfn) + elog(ERROR, "finalfunc of aggpartialfunc must match serialfunc of aggregate when stype is internal"); + } + else if (aggpartialform->aggfinalfn != InvalidOid) + elog(ERROR, "finalfunc of aggpartialfunc must not be supplied when stype isn't internal"); + + ReleaseSysCache(aggtup); + } + } + /* handle sortop, if supplied */ if (aggsortopName) { @@ -684,6 +775,9 @@ AggregateCreate(const char *aggName, values[Anum_pg_aggregate_aggminitval - 1] = CStringGetTextDatum(aggminitval); else nulls[Anum_pg_aggregate_aggminitval - 1] = true; + if (isaggpartialfn) + aggpartialfn = procOid; + values[Anum_pg_aggregate_aggpartialfn - 1] = ObjectIdGetDatum(aggpartialfn); if (replace) oldtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(procOid)); @@ -805,6 +899,13 @@ AggregateCreate(const char *aggName, add_exact_object_address(&referenced, addrs); } + /* Depends on aggpartialfunc, if any */ + if (OidIsValid(aggpartialfn) && (aggpartialfn != procOid)) + { + ObjectAddressSet(referenced, ProcedureRelationId, aggpartialfn); + add_exact_object_address(&referenced, addrs); + } + record_object_address_dependencies(&myself, addrs, DEPENDENCY_NORMAL); free_object_addresses(addrs); return myself; diff --git a/src/backend/commands/aggregatecmds.c b/src/backend/commands/aggregatecmds.c index fda9d1aa77..8c8f1789e9 100644 --- a/src/backend/commands/aggregatecmds.c +++ b/src/backend/commands/aggregatecmds.c @@ -70,6 +70,7 @@ DefineAggregate(ParseState *pstate, List *combinefuncName = NIL; List *serialfuncName = NIL; List *deserialfuncName = NIL; + List *aggpartialfuncName = NIL; List *mtransfuncName = NIL; List *minvtransfuncName = NIL; List *mfinalfuncName = NIL; @@ -143,6 +144,8 @@ DefineAggregate(ParseState *pstate, serialfuncName = defGetQualifiedName(defel); else if (strcmp(defel->defname, "deserialfunc") == 0) deserialfuncName = defGetQualifiedName(defel); + else if (strcmp(defel->defname, "aggpartialfunc") == 0) + aggpartialfuncName = defGetQualifiedName(defel); else if (strcmp(defel->defname, "msfunc") == 0) mtransfuncName = defGetQualifiedName(defel); else if (strcmp(defel->defname, "minvfunc") == 0) @@ -461,6 +464,7 @@ DefineAggregate(ParseState *pstate, mtransfuncName, /* fwd trans function name */ minvtransfuncName, /* inv trans function name */ mfinalfuncName, /* final function name */ + aggpartialfuncName, finalfuncExtraArgs, mfinalfuncExtraArgs, finalfuncModify, diff --git a/src/backend/optimizer/plan/planner.c b/src/backend/optimizer/plan/planner.c index 1e4dd27dba..f4b8781e1e 100644 --- a/src/backend/optimizer/plan/planner.c +++ b/src/backend/optimizer/plan/planner.c @@ -206,7 +206,7 @@ static PathTarget *make_group_input_target(PlannerInfo *root, PathTarget *final_target); static PathTarget *make_partial_grouping_target(PlannerInfo *root, PathTarget *grouping_target, - Node *havingQual); + GroupPathExtraData *extra); static List *postprocess_setop_tlist(List *new_tlist, List *orig_tlist); static void optimize_window_clauses(PlannerInfo *root, WindowFuncLists *wflists); @@ -5398,6 +5398,64 @@ make_group_input_target(PlannerInfo *root, PathTarget *final_target) return set_pathtarget_cost_width(root, input_target); } +/* + * setGroupClausePartial + * Generate a groupClause for partial aggregate and set it to GroupPathExtraData. + */ +static void +setGroupClausePartial(PathTarget *partial_target, List *non_group_exprs, + List *groupClause, GroupPathExtraData *extra) +{ + int exprno, + refno; + ListCell *lc; + Index maxRef = 0; + List *exprs_processed = NIL; + + foreach(lc, groupClause) + { + SortGroupClause *sgc = (SortGroupClause *) lfirst(lc); + + if (sgc->tleSortGroupRef > maxRef) + maxRef = sgc->tleSortGroupRef; + } + maxRef++; + + extra->groupClausePartial = list_copy_deep(groupClause); + extra->partial_target = copy_pathtarget(partial_target); + + if (!partial_target->exprs) + return; + + foreach(lc, non_group_exprs) + { + Expr *expr = (Expr *) lfirst(lc); + + refno = -1; + if (list_member(exprs_processed, expr) || !IsA(expr, Var)) + continue; + exprs_processed = lappend(exprs_processed, expr); + for (exprno = 0; exprno < partial_target->exprs->length; exprno++) + { + Expr *target_expr = (Expr *) list_nth(partial_target->exprs, exprno); + + if (equal(target_expr, expr)) + { + refno = exprno; + break; + } + } + if (refno < 0) + { + SortGroupClause *grpcl = makeNode(SortGroupClause); + + grpcl->tleSortGroupRef = maxRef++; + extra->groupClausePartial = lappend(extra->groupClausePartial, grpcl); + add_column_to_pathtarget(extra->partial_target, expr, grpcl->tleSortGroupRef); + } + } +} + /* * make_partial_grouping_target * Generate appropriate PathTarget for output of partial aggregate @@ -5417,7 +5473,7 @@ make_group_input_target(PlannerInfo *root, PathTarget *final_target) static PathTarget * make_partial_grouping_target(PlannerInfo *root, PathTarget *grouping_target, - Node *havingQual) + GroupPathExtraData *extra) { PathTarget *partial_target; List *non_group_cols; @@ -5459,8 +5515,8 @@ make_partial_grouping_target(PlannerInfo *root, /* * If there's a HAVING clause, we'll need the Vars/Aggrefs it uses, too. */ - if (havingQual) - non_group_cols = lappend(non_group_cols, havingQual); + if (extra->havingQual) + non_group_cols = lappend(non_group_cols, extra->havingQual); /* * Pull out all the Vars, PlaceHolderVars, and Aggrefs mentioned in @@ -5473,7 +5529,7 @@ make_partial_grouping_target(PlannerInfo *root, PVC_INCLUDE_AGGREGATES | PVC_RECURSE_WINDOWFUNCS | PVC_INCLUDE_PLACEHOLDERS); - + setGroupClausePartial(partial_target, non_group_exprs, root->parse->groupClause, extra); add_new_columns_to_pathtarget(partial_target, non_group_exprs); /* @@ -5502,6 +5558,7 @@ make_partial_grouping_target(PlannerInfo *root, lfirst(lc) = newaggref; } } + extra->partial_target->exprs = partial_target->exprs; /* clean up cruft */ list_free(non_group_exprs); @@ -7096,7 +7153,7 @@ create_partial_grouping_paths(PlannerInfo *root, */ partially_grouped_rel->reltarget = make_partial_grouping_target(root, grouped_rel->reltarget, - extra->havingQual); + extra); if (!extra->partial_costs_set) { diff --git a/src/bin/pg_dump/pg_dump.c b/src/bin/pg_dump/pg_dump.c index 3af97a6039..5ef3024072 100644 --- a/src/bin/pg_dump/pg_dump.c +++ b/src/bin/pg_dump/pg_dump.c @@ -13676,6 +13676,7 @@ dumpAgg(Archive *fout, const AggInfo *agginfo) const char *aggmtransfn; const char *aggminvtransfn; const char *aggmfinalfn; + const char *aggpartialfn; bool aggfinalextra; bool aggmfinalextra; char aggfinalmodify; @@ -13758,11 +13759,11 @@ dumpAgg(Archive *fout, const AggInfo *agginfo) if (fout->remoteVersion >= 110000) appendPQExpBufferStr(query, "aggfinalmodify,\n" - "aggmfinalmodify\n"); + "aggmfinalmodify,\n"); else appendPQExpBufferStr(query, "'0' AS aggfinalmodify,\n" - "'0' AS aggmfinalmodify\n"); + "'0' AS aggmfinalmodify,\n"); appendPQExpBufferStr(query, "FROM pg_catalog.pg_aggregate a, pg_catalog.pg_proc p " @@ -13788,6 +13789,7 @@ dumpAgg(Archive *fout, const AggInfo *agginfo) aggcombinefn = PQgetvalue(res, 0, PQfnumber(res, "aggcombinefn")); aggserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggserialfn")); aggdeserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggdeserialfn")); + aggpartialfn = PQgetvalue(res, 0, PQfnumber(res, "aggpartialfn")); aggmtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggmtransfn")); aggminvtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggminvtransfn")); aggmfinalfn = PQgetvalue(res, 0, PQfnumber(res, "aggmfinalfn")); @@ -13877,6 +13879,9 @@ dumpAgg(Archive *fout, const AggInfo *agginfo) if (strcmp(aggdeserialfn, "-") != 0) appendPQExpBuffer(details, ",\n DESERIALFUNC = %s", aggdeserialfn); + if (strcmp(aggpartialfn, "-") != 0) + appendPQExpBuffer(details, ",\n AGGPARTIALFUNC = %s", aggpartialfn); + if (strcmp(aggmtransfn, "-") != 0) { appendPQExpBuffer(details, ",\n MSFUNC = %s,\n MINVFUNC = %s,\n MSTYPE = %s", diff --git a/src/include/catalog/pg_aggregate.dat b/src/include/catalog/pg_aggregate.dat index 1bc1d97d74..dd97b4dc34 100644 --- a/src/include/catalog/pg_aggregate.dat +++ b/src/include/catalog/pg_aggregate.dat @@ -13,22 +13,44 @@ [ # avg +{ aggfnoid => 'avg_p_int8', aggtransfn => 'int8_avg_accum', + aggfinalfn => 'int8_avg_serialize', aggcombinefn => 'int8_avg_combine', + aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize', + aggtranstype => 'internal', aggtransspace => '48', + aggpartialfn => 'avg_p_int8' }, { aggfnoid => 'avg(int8)', aggtransfn => 'int8_avg_accum', aggfinalfn => 'numeric_poly_avg', aggcombinefn => 'int8_avg_combine', aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize', aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv', aggmfinalfn => 'numeric_poly_avg', aggtranstype => 'internal', - aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' }, + aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'avg_p_int8' }, +{ aggfnoid => 'avg_p_int4', aggtransfn => 'int4_avg_accum', + aggcombinefn => 'int4_avg_combine', aggtranstype => '_int8', + agginitval => '{0,0}', + aggpartialfn => 'avg_p_int4' }, { aggfnoid => 'avg(int4)', aggtransfn => 'int4_avg_accum', aggfinalfn => 'int8_avg', aggcombinefn => 'int4_avg_combine', aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv', aggmfinalfn => 'int8_avg', aggtranstype => '_int8', aggmtranstype => '_int8', - agginitval => '{0,0}', aggminitval => '{0,0}' }, + agginitval => '{0,0}', aggminitval => '{0,0}', + aggpartialfn => 'avg_p_int4' }, +{ aggfnoid => 'avg_p_int2', aggtransfn => 'int2_avg_accum', + aggcombinefn => 'int4_avg_combine', aggtranstype => '_int8', + agginitval => '{0,0}', + aggpartialfn => 'avg_p_int2' }, { aggfnoid => 'avg(int2)', aggtransfn => 'int2_avg_accum', aggfinalfn => 'int8_avg', aggcombinefn => 'int4_avg_combine', aggmtransfn => 'int2_avg_accum', aggminvtransfn => 'int2_avg_accum_inv', aggmfinalfn => 'int8_avg', aggtranstype => '_int8', aggmtranstype => '_int8', - agginitval => '{0,0}', aggminitval => '{0,0}' }, + agginitval => '{0,0}', aggminitval => '{0,0}', + aggpartialfn => 'avg_p_int2' }, +{ aggfnoid => 'avg_p_numeric', aggtransfn => 'numeric_avg_accum', + aggfinalfn => 'numeric_avg_serialize', aggcombinefn => 'numeric_avg_combine', + aggserialfn => 'numeric_avg_serialize', + aggdeserialfn => 'numeric_avg_deserialize', aggtranstype => 'internal', + aggtransspace => '128', + aggpartialfn => 'avg_p_numeric' }, { aggfnoid => 'avg(numeric)', aggtransfn => 'numeric_avg_accum', aggfinalfn => 'numeric_avg', aggcombinefn => 'numeric_avg_combine', aggserialfn => 'numeric_avg_serialize', @@ -36,46 +58,80 @@ aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv', aggmfinalfn => 'numeric_avg', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'avg_p_numeric' }, +{ aggfnoid => 'avg_p_float4', aggtransfn => 'float4_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'avg_p_float4' }, { aggfnoid => 'avg(float4)', aggtransfn => 'float4_accum', aggfinalfn => 'float8_avg', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'avg_p_float4' }, +{ aggfnoid => 'avg_p_float8', aggtransfn => 'float8_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'avg_p_float8' }, { aggfnoid => 'avg(float8)', aggtransfn => 'float8_accum', aggfinalfn => 'float8_avg', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'avg_p_float8' }, +{ aggfnoid => 'avg_p_interval', aggtransfn => 'interval_accum', + aggcombinefn => 'interval_combine', aggtranstype => '_interval', + agginitval => '{0 second,0 second}', + aggpartialfn => 'avg_p_interval' }, { aggfnoid => 'avg(interval)', aggtransfn => 'interval_accum', aggfinalfn => 'interval_avg', aggcombinefn => 'interval_combine', aggmtransfn => 'interval_accum', aggminvtransfn => 'interval_accum_inv', aggmfinalfn => 'interval_avg', aggtranstype => '_interval', aggmtranstype => '_interval', agginitval => '{0 second,0 second}', - aggminitval => '{0 second,0 second}' }, + aggminitval => '{0 second,0 second}', + aggpartialfn => 'avg_p_interval' }, # sum +{ aggfnoid => 'sum_p_int8', aggtransfn => 'int8_avg_accum', + aggfinalfn => 'int8_avg_serialize', aggcombinefn => 'int8_avg_combine', + aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize', + aggtranstype => 'internal', aggtransspace => '48', + aggpartialfn => 'sum_p_int8' }, { aggfnoid => 'sum(int8)', aggtransfn => 'int8_avg_accum', aggfinalfn => 'numeric_poly_sum', aggcombinefn => 'int8_avg_combine', aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize', aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv', aggmfinalfn => 'numeric_poly_sum', aggtranstype => 'internal', - aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' }, + aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'sum_p_int8' }, { aggfnoid => 'sum(int4)', aggtransfn => 'int4_sum', aggcombinefn => 'int8pl', aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv', aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8', - aggmtranstype => '_int8', aggminitval => '{0,0}' }, + aggmtranstype => '_int8', aggminitval => '{0,0}', + aggpartialfn => 'sum(int4)' }, { aggfnoid => 'sum(int2)', aggtransfn => 'int2_sum', aggcombinefn => 'int8pl', aggmtransfn => 'int2_avg_accum', aggminvtransfn => 'int2_avg_accum_inv', aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8', - aggmtranstype => '_int8', aggminitval => '{0,0}' }, + aggmtranstype => '_int8', aggminitval => '{0,0}', + aggpartialfn => 'sum(int2)' }, { aggfnoid => 'sum(float4)', aggtransfn => 'float4pl', - aggcombinefn => 'float4pl', aggtranstype => 'float4' }, + aggcombinefn => 'float4pl', aggtranstype => 'float4', + aggpartialfn => 'sum(float4)' }, { aggfnoid => 'sum(float8)', aggtransfn => 'float8pl', - aggcombinefn => 'float8pl', aggtranstype => 'float8' }, + aggcombinefn => 'float8pl', aggtranstype => 'float8', + aggpartialfn => 'sum(float8)' }, { aggfnoid => 'sum(money)', aggtransfn => 'cash_pl', aggcombinefn => 'cash_pl', aggmtransfn => 'cash_pl', aggminvtransfn => 'cash_mi', - aggtranstype => 'money', aggmtranstype => 'money' }, + aggtranstype => 'money', aggmtranstype => 'money', + aggpartialfn => 'sum(money)' }, { aggfnoid => 'sum(interval)', aggtransfn => 'interval_pl', aggcombinefn => 'interval_pl', aggmtransfn => 'interval_pl', aggminvtransfn => 'interval_mi', aggtranstype => 'interval', - aggmtranstype => 'interval' }, + aggmtranstype => 'interval', + aggpartialfn => 'sum(interval)' }, +{ aggfnoid => 'sum_p_numeric', aggtransfn => 'numeric_avg_accum', + aggfinalfn => 'numeric_avg_serialize', aggcombinefn => 'numeric_avg_combine', + aggserialfn => 'numeric_avg_serialize', + aggdeserialfn => 'numeric_avg_deserialize', aggtranstype => 'internal', + aggtransspace => '128', + aggpartialfn => 'sum_p_numeric' }, { aggfnoid => 'sum(numeric)', aggtransfn => 'numeric_avg_accum', aggfinalfn => 'numeric_sum', aggcombinefn => 'numeric_avg_combine', aggserialfn => 'numeric_avg_serialize', @@ -83,269 +139,436 @@ aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv', aggmfinalfn => 'numeric_sum', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'sum_p_numeric' }, # max { aggfnoid => 'max(int8)', aggtransfn => 'int8larger', aggcombinefn => 'int8larger', aggsortop => '>(int8,int8)', - aggtranstype => 'int8' }, + aggtranstype => 'int8', + aggpartialfn => 'max(int8)' }, { aggfnoid => 'max(int4)', aggtransfn => 'int4larger', aggcombinefn => 'int4larger', aggsortop => '>(int4,int4)', - aggtranstype => 'int4' }, + aggtranstype => 'int4', + aggpartialfn => 'max(int4)' }, { aggfnoid => 'max(int2)', aggtransfn => 'int2larger', aggcombinefn => 'int2larger', aggsortop => '>(int2,int2)', - aggtranstype => 'int2' }, + aggtranstype => 'int2', + aggpartialfn => 'max(int2)' }, { aggfnoid => 'max(oid)', aggtransfn => 'oidlarger', aggcombinefn => 'oidlarger', aggsortop => '>(oid,oid)', - aggtranstype => 'oid' }, + aggtranstype => 'oid', + aggpartialfn => 'max(oid)' }, { aggfnoid => 'max(float4)', aggtransfn => 'float4larger', aggcombinefn => 'float4larger', aggsortop => '>(float4,float4)', - aggtranstype => 'float4' }, + aggtranstype => 'float4', + aggpartialfn => 'max(float4)' }, { aggfnoid => 'max(float8)', aggtransfn => 'float8larger', aggcombinefn => 'float8larger', aggsortop => '>(float8,float8)', - aggtranstype => 'float8' }, + aggtranstype => 'float8', + aggpartialfn => 'max(float8)' }, { aggfnoid => 'max(date)', aggtransfn => 'date_larger', aggcombinefn => 'date_larger', aggsortop => '>(date,date)', - aggtranstype => 'date' }, + aggtranstype => 'date', + aggpartialfn => 'max(date)' }, { aggfnoid => 'max(time)', aggtransfn => 'time_larger', aggcombinefn => 'time_larger', aggsortop => '>(time,time)', - aggtranstype => 'time' }, + aggtranstype => 'time', + aggpartialfn => 'max(time)' }, { aggfnoid => 'max(timetz)', aggtransfn => 'timetz_larger', aggcombinefn => 'timetz_larger', aggsortop => '>(timetz,timetz)', - aggtranstype => 'timetz' }, + aggtranstype => 'timetz', + aggpartialfn => 'max(timetz)' }, { aggfnoid => 'max(money)', aggtransfn => 'cashlarger', aggcombinefn => 'cashlarger', aggsortop => '>(money,money)', - aggtranstype => 'money' }, + aggtranstype => 'money', + aggpartialfn => 'max(money)' }, { aggfnoid => 'max(timestamp)', aggtransfn => 'timestamp_larger', aggcombinefn => 'timestamp_larger', aggsortop => '>(timestamp,timestamp)', - aggtranstype => 'timestamp' }, + aggtranstype => 'timestamp', + aggpartialfn => 'max(timestamp)' }, { aggfnoid => 'max(timestamptz)', aggtransfn => 'timestamptz_larger', aggcombinefn => 'timestamptz_larger', - aggsortop => '>(timestamptz,timestamptz)', aggtranstype => 'timestamptz' }, + aggsortop => '>(timestamptz,timestamptz)', aggtranstype => 'timestamptz', + aggpartialfn => 'max(timestamptz)' }, { aggfnoid => 'max(interval)', aggtransfn => 'interval_larger', aggcombinefn => 'interval_larger', aggsortop => '>(interval,interval)', - aggtranstype => 'interval' }, + aggtranstype => 'interval', + aggpartialfn => 'max(interval)' }, { aggfnoid => 'max(text)', aggtransfn => 'text_larger', aggcombinefn => 'text_larger', aggsortop => '>(text,text)', - aggtranstype => 'text' }, + aggtranstype => 'text', + aggpartialfn => 'max(text)' }, { aggfnoid => 'max(numeric)', aggtransfn => 'numeric_larger', aggcombinefn => 'numeric_larger', aggsortop => '>(numeric,numeric)', - aggtranstype => 'numeric' }, + aggtranstype => 'numeric', + aggpartialfn => 'max(numeric)' }, { aggfnoid => 'max(anyarray)', aggtransfn => 'array_larger', aggcombinefn => 'array_larger', aggsortop => '>(anyarray,anyarray)', - aggtranstype => 'anyarray' }, + aggtranstype => 'anyarray', + aggpartialfn => 'max(anyarray)' }, { aggfnoid => 'max(bpchar)', aggtransfn => 'bpchar_larger', aggcombinefn => 'bpchar_larger', aggsortop => '>(bpchar,bpchar)', - aggtranstype => 'bpchar' }, + aggtranstype => 'bpchar', + aggpartialfn => 'max(bpchar)' }, { aggfnoid => 'max(tid)', aggtransfn => 'tidlarger', aggcombinefn => 'tidlarger', aggsortop => '>(tid,tid)', - aggtranstype => 'tid' }, + aggtranstype => 'tid', + aggpartialfn => 'max(tid)' }, { aggfnoid => 'max(anyenum)', aggtransfn => 'enum_larger', aggcombinefn => 'enum_larger', aggsortop => '>(anyenum,anyenum)', - aggtranstype => 'anyenum' }, + aggtranstype => 'anyenum', + aggpartialfn => 'max(anyenum)' }, { aggfnoid => 'max(inet)', aggtransfn => 'network_larger', aggcombinefn => 'network_larger', aggsortop => '>(inet,inet)', - aggtranstype => 'inet' }, + aggtranstype => 'inet', + aggpartialfn => 'max(inet)' }, { aggfnoid => 'max(pg_lsn)', aggtransfn => 'pg_lsn_larger', aggcombinefn => 'pg_lsn_larger', aggsortop => '>(pg_lsn,pg_lsn)', - aggtranstype => 'pg_lsn' }, + aggtranstype => 'pg_lsn', + aggpartialfn => 'max(pg_lsn)' }, { aggfnoid => 'max(xid8)', aggtransfn => 'xid8_larger', aggcombinefn => 'xid8_larger', aggsortop => '>(xid8,xid8)', - aggtranstype => 'xid8' }, + aggtranstype => 'xid8', + aggpartialfn => 'max(xid8)' }, # min { aggfnoid => 'min(int8)', aggtransfn => 'int8smaller', aggcombinefn => 'int8smaller', aggsortop => '<(int8,int8)', - aggtranstype => 'int8' }, + aggtranstype => 'int8', + aggpartialfn => 'min(int8)' }, { aggfnoid => 'min(int4)', aggtransfn => 'int4smaller', aggcombinefn => 'int4smaller', aggsortop => '<(int4,int4)', - aggtranstype => 'int4' }, + aggtranstype => 'int4', + aggpartialfn => 'min(int4)' }, { aggfnoid => 'min(int2)', aggtransfn => 'int2smaller', aggcombinefn => 'int2smaller', aggsortop => '<(int2,int2)', - aggtranstype => 'int2' }, + aggtranstype => 'int2', + aggpartialfn => 'min(int2)' }, { aggfnoid => 'min(oid)', aggtransfn => 'oidsmaller', aggcombinefn => 'oidsmaller', aggsortop => '<(oid,oid)', - aggtranstype => 'oid' }, + aggtranstype => 'oid', + aggpartialfn => 'min(oid)' }, { aggfnoid => 'min(float4)', aggtransfn => 'float4smaller', aggcombinefn => 'float4smaller', aggsortop => '<(float4,float4)', - aggtranstype => 'float4' }, + aggtranstype => 'float4', + aggpartialfn => 'min(float4)' }, { aggfnoid => 'min(float8)', aggtransfn => 'float8smaller', aggcombinefn => 'float8smaller', aggsortop => '<(float8,float8)', - aggtranstype => 'float8' }, + aggtranstype => 'float8', + aggpartialfn => 'min(float8)' }, { aggfnoid => 'min(date)', aggtransfn => 'date_smaller', aggcombinefn => 'date_smaller', aggsortop => '<(date,date)', - aggtranstype => 'date' }, + aggtranstype => 'date', + aggpartialfn => 'min(date)' }, { aggfnoid => 'min(time)', aggtransfn => 'time_smaller', aggcombinefn => 'time_smaller', aggsortop => '<(time,time)', - aggtranstype => 'time' }, + aggtranstype => 'time', + aggpartialfn => 'min(time)' }, { aggfnoid => 'min(timetz)', aggtransfn => 'timetz_smaller', aggcombinefn => 'timetz_smaller', aggsortop => '<(timetz,timetz)', - aggtranstype => 'timetz' }, + aggtranstype => 'timetz', + aggpartialfn => 'min(timetz)' }, { aggfnoid => 'min(money)', aggtransfn => 'cashsmaller', aggcombinefn => 'cashsmaller', aggsortop => '<(money,money)', - aggtranstype => 'money' }, + aggtranstype => 'money', + aggpartialfn => 'min(money)' }, { aggfnoid => 'min(timestamp)', aggtransfn => 'timestamp_smaller', aggcombinefn => 'timestamp_smaller', aggsortop => '<(timestamp,timestamp)', - aggtranstype => 'timestamp' }, + aggtranstype => 'timestamp', + aggpartialfn => 'min(timestamp)' }, { aggfnoid => 'min(timestamptz)', aggtransfn => 'timestamptz_smaller', aggcombinefn => 'timestamptz_smaller', - aggsortop => '<(timestamptz,timestamptz)', aggtranstype => 'timestamptz' }, + aggsortop => '<(timestamptz,timestamptz)', aggtranstype => 'timestamptz', + aggpartialfn => 'min(timestamptz)' }, { aggfnoid => 'min(interval)', aggtransfn => 'interval_smaller', aggcombinefn => 'interval_smaller', aggsortop => '<(interval,interval)', - aggtranstype => 'interval' }, + aggtranstype => 'interval', + aggpartialfn => 'min(interval)' }, { aggfnoid => 'min(text)', aggtransfn => 'text_smaller', aggcombinefn => 'text_smaller', aggsortop => '<(text,text)', - aggtranstype => 'text' }, + aggtranstype => 'text', + aggpartialfn => 'min(text)' }, { aggfnoid => 'min(numeric)', aggtransfn => 'numeric_smaller', aggcombinefn => 'numeric_smaller', aggsortop => '<(numeric,numeric)', - aggtranstype => 'numeric' }, + aggtranstype => 'numeric', + aggpartialfn => 'min(numeric)' }, { aggfnoid => 'min(anyarray)', aggtransfn => 'array_smaller', aggcombinefn => 'array_smaller', aggsortop => '<(anyarray,anyarray)', - aggtranstype => 'anyarray' }, + aggtranstype => 'anyarray', + aggpartialfn => 'min(anyarray)' }, { aggfnoid => 'min(bpchar)', aggtransfn => 'bpchar_smaller', aggcombinefn => 'bpchar_smaller', aggsortop => '<(bpchar,bpchar)', - aggtranstype => 'bpchar' }, + aggtranstype => 'bpchar', + aggpartialfn => 'min(bpchar)' }, { aggfnoid => 'min(tid)', aggtransfn => 'tidsmaller', aggcombinefn => 'tidsmaller', aggsortop => '<(tid,tid)', - aggtranstype => 'tid' }, + aggtranstype => 'tid', + aggpartialfn => 'min(tid)' }, { aggfnoid => 'min(anyenum)', aggtransfn => 'enum_smaller', aggcombinefn => 'enum_smaller', aggsortop => '<(anyenum,anyenum)', - aggtranstype => 'anyenum' }, + aggtranstype => 'anyenum', + aggpartialfn => 'min(anyenum)' }, { aggfnoid => 'min(inet)', aggtransfn => 'network_smaller', aggcombinefn => 'network_smaller', aggsortop => '<(inet,inet)', - aggtranstype => 'inet' }, + aggtranstype => 'inet', + aggpartialfn => 'min(inet)' }, { aggfnoid => 'min(pg_lsn)', aggtransfn => 'pg_lsn_smaller', aggcombinefn => 'pg_lsn_smaller', aggsortop => '<(pg_lsn,pg_lsn)', - aggtranstype => 'pg_lsn' }, + aggtranstype => 'pg_lsn', + aggpartialfn => 'min(pg_lsn)' }, { aggfnoid => 'min(xid8)', aggtransfn => 'xid8_smaller', aggcombinefn => 'xid8_smaller', aggsortop => '<(xid8,xid8)', - aggtranstype => 'xid8' }, + aggtranstype => 'xid8', + aggpartialfn => 'min(xid8)' }, # count { aggfnoid => 'count(any)', aggtransfn => 'int8inc_any', aggcombinefn => 'int8pl', aggmtransfn => 'int8inc_any', aggminvtransfn => 'int8dec_any', aggtranstype => 'int8', - aggmtranstype => 'int8', agginitval => '0', aggminitval => '0' }, + aggmtranstype => 'int8', agginitval => '0', aggminitval => '0', + aggpartialfn => 'count(any)' }, { aggfnoid => 'count()', aggtransfn => 'int8inc', aggcombinefn => 'int8pl', aggmtransfn => 'int8inc', aggminvtransfn => 'int8dec', aggtranstype => 'int8', - aggmtranstype => 'int8', agginitval => '0', aggminitval => '0' }, + aggmtranstype => 'int8', agginitval => '0', aggminitval => '0', + aggpartialfn => 'count()' }, # var_pop +{ aggfnoid => 'var_pop_p_int8', aggtransfn => 'int8_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'var_pop_p_int8' }, { aggfnoid => 'var_pop(int8)', aggtransfn => 'int8_accum', aggfinalfn => 'numeric_var_pop', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'int8_accum', aggminvtransfn => 'int8_accum_inv', aggmfinalfn => 'numeric_var_pop', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'var_pop_p_int8' }, +{ aggfnoid => 'var_pop_p_int4', aggtransfn => 'int4_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'var_pop_p_int4' }, { aggfnoid => 'var_pop(int4)', aggtransfn => 'int4_accum', aggfinalfn => 'numeric_poly_var_pop', aggcombinefn => 'numeric_poly_combine', aggserialfn => 'numeric_poly_serialize', aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int4_accum', aggminvtransfn => 'int4_accum_inv', aggmfinalfn => 'numeric_poly_var_pop', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'var_pop_p_int4' }, +{ aggfnoid => 'var_pop_p_int2', aggtransfn => 'int2_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'var_pop_p_int2' }, { aggfnoid => 'var_pop(int2)', aggtransfn => 'int2_accum', aggfinalfn => 'numeric_poly_var_pop', aggcombinefn => 'numeric_poly_combine', aggserialfn => 'numeric_poly_serialize', aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int2_accum', aggminvtransfn => 'int2_accum_inv', aggmfinalfn => 'numeric_poly_var_pop', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'var_pop_p_int2' }, +{ aggfnoid => 'var_pop_p_float4', aggtransfn => 'float4_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'var_pop_p_float4' }, { aggfnoid => 'var_pop(float4)', aggtransfn => 'float4_accum', aggfinalfn => 'float8_var_pop', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'var_pop_p_float4' }, +{ aggfnoid => 'var_pop_p_float8', aggtransfn => 'float8_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'var_pop_p_float8' }, { aggfnoid => 'var_pop(float8)', aggtransfn => 'float8_accum', aggfinalfn => 'float8_var_pop', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'var_pop_p_float8' }, +{ aggfnoid => 'var_pop_p_numeric', aggtransfn => 'numeric_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'var_pop_p_numeric' }, { aggfnoid => 'var_pop(numeric)', aggtransfn => 'numeric_accum', aggfinalfn => 'numeric_var_pop', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'numeric_accum', aggminvtransfn => 'numeric_accum_inv', aggmfinalfn => 'numeric_var_pop', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'var_pop_p_numeric' }, # var_samp +{ aggfnoid => 'var_samp_p_int8', aggtransfn => 'int8_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'var_samp_p_int8' }, { aggfnoid => 'var_samp(int8)', aggtransfn => 'int8_accum', aggfinalfn => 'numeric_var_samp', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'int8_accum', aggminvtransfn => 'int8_accum_inv', aggmfinalfn => 'numeric_var_samp', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'var_samp_p_int8' }, +{ aggfnoid => 'var_samp_p_int4', aggtransfn => 'int4_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'var_samp_p_int4' }, { aggfnoid => 'var_samp(int4)', aggtransfn => 'int4_accum', aggfinalfn => 'numeric_poly_var_samp', aggcombinefn => 'numeric_poly_combine', aggserialfn => 'numeric_poly_serialize', aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int4_accum', aggminvtransfn => 'int4_accum_inv', aggmfinalfn => 'numeric_poly_var_samp', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'var_samp_p_int4' }, +{ aggfnoid => 'var_samp_p_int2', aggtransfn => 'int2_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'var_samp_p_int2' }, { aggfnoid => 'var_samp(int2)', aggtransfn => 'int2_accum', aggfinalfn => 'numeric_poly_var_samp', aggcombinefn => 'numeric_poly_combine', aggserialfn => 'numeric_poly_serialize', aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int2_accum', aggminvtransfn => 'int2_accum_inv', aggmfinalfn => 'numeric_poly_var_samp', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'var_samp_p_int2' }, +{ aggfnoid => 'var_samp_p_float4', aggtransfn => 'float4_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'var_samp_p_float4' }, { aggfnoid => 'var_samp(float4)', aggtransfn => 'float4_accum', aggfinalfn => 'float8_var_samp', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'var_samp_p_float4' }, +{ aggfnoid => 'var_samp_p_float8', aggtransfn => 'float8_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'var_samp_p_float8' }, { aggfnoid => 'var_samp(float8)', aggtransfn => 'float8_accum', aggfinalfn => 'float8_var_samp', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'var_samp_p_float8' }, +{ aggfnoid => 'var_samp_p_numeric', aggtransfn => 'numeric_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'var_samp_p_numeric' }, { aggfnoid => 'var_samp(numeric)', aggtransfn => 'numeric_accum', aggfinalfn => 'numeric_var_samp', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'numeric_accum', aggminvtransfn => 'numeric_accum_inv', aggmfinalfn => 'numeric_var_samp', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'var_samp_p_numeric' }, # variance: historical Postgres syntax for var_samp +{ aggfnoid => 'variance_p_int8', aggtransfn => 'int8_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'variance_p_int8' }, { aggfnoid => 'variance(int8)', aggtransfn => 'int8_accum', aggfinalfn => 'numeric_var_samp', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'int8_accum', aggminvtransfn => 'int8_accum_inv', aggmfinalfn => 'numeric_var_samp', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'variance_p_int8' }, +{ aggfnoid => 'variance_p_int4', aggtransfn => 'int4_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'variance_p_int4' }, { aggfnoid => 'variance(int4)', aggtransfn => 'int4_accum', aggfinalfn => 'numeric_poly_var_samp', aggcombinefn => 'numeric_poly_combine', aggserialfn => 'numeric_poly_serialize', aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int4_accum', aggminvtransfn => 'int4_accum_inv', aggmfinalfn => 'numeric_poly_var_samp', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'variance_p_int4' }, +{ aggfnoid => 'variance_p_int2', aggtransfn => 'int2_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'variance_p_int2' }, { aggfnoid => 'variance(int2)', aggtransfn => 'int2_accum', aggfinalfn => 'numeric_poly_var_samp', aggcombinefn => 'numeric_poly_combine', aggserialfn => 'numeric_poly_serialize', aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int2_accum', aggminvtransfn => 'int2_accum_inv', aggmfinalfn => 'numeric_poly_var_samp', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'variance_p_int2' }, +{ aggfnoid => 'variance_p_float4', aggtransfn => 'float4_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'variance_p_float4' }, { aggfnoid => 'variance(float4)', aggtransfn => 'float4_accum', aggfinalfn => 'float8_var_samp', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'variance_p_float4' }, +{ aggfnoid => 'variance_p_float8', aggtransfn => 'float8_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'variance_p_float8' }, { aggfnoid => 'variance(float8)', aggtransfn => 'float8_accum', aggfinalfn => 'float8_var_samp', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'variance_p_float8' }, +{ aggfnoid => 'variance_p_numeric', aggtransfn => 'numeric_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'variance_p_numeric' }, { aggfnoid => 'variance(numeric)', aggtransfn => 'numeric_accum', aggfinalfn => 'numeric_var_samp', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'numeric_accum', aggminvtransfn => 'numeric_accum_inv', aggmfinalfn => 'numeric_var_samp', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'variance_p_numeric' }, # stddev_pop +{ aggfnoid => 'stddev_pop_p_int8', aggtransfn => 'int8_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'stddev_pop_p_int8' }, { aggfnoid => 'stddev_pop(int8)', aggtransfn => 'int8_accum', aggfinalfn => 'numeric_stddev_pop', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'int8_accum', aggminvtransfn => 'int8_accum_inv', aggmfinalfn => 'numeric_stddev_pop', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'stddev_pop_p_int8' }, +{ aggfnoid => 'stddev_pop_p_int4', aggtransfn => 'int4_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'stddev_pop_p_int4' }, { aggfnoid => 'stddev_pop(int4)', aggtransfn => 'int4_accum', aggfinalfn => 'numeric_poly_stddev_pop', aggcombinefn => 'numeric_poly_combine', @@ -353,7 +576,14 @@ aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int4_accum', aggminvtransfn => 'int4_accum_inv', aggmfinalfn => 'numeric_poly_stddev_pop', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'stddev_pop_p_int4' }, +{ aggfnoid => 'stddev_pop_p_int2', aggtransfn => 'int2_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'stddev_pop_p_int2' }, { aggfnoid => 'stddev_pop(int2)', aggtransfn => 'int2_accum', aggfinalfn => 'numeric_poly_stddev_pop', aggcombinefn => 'numeric_poly_combine', @@ -361,29 +591,58 @@ aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int2_accum', aggminvtransfn => 'int2_accum_inv', aggmfinalfn => 'numeric_poly_stddev_pop', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'stddev_pop_p_int2' }, +{ aggfnoid => 'stddev_pop_p_float4', aggtransfn => 'float4_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'stddev_pop_p_float4' }, { aggfnoid => 'stddev_pop(float4)', aggtransfn => 'float4_accum', aggfinalfn => 'float8_stddev_pop', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'stddev_pop_p_float4' }, +{ aggfnoid => 'stddev_pop_p_float8', aggtransfn => 'float8_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'stddev_pop_p_float8' }, { aggfnoid => 'stddev_pop(float8)', aggtransfn => 'float8_accum', aggfinalfn => 'float8_stddev_pop', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'stddev_pop_p_float8' }, +{ aggfnoid => 'stddev_pop_p_numeric', aggtransfn => 'numeric_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'stddev_pop_p_numeric' }, { aggfnoid => 'stddev_pop(numeric)', aggtransfn => 'numeric_accum', aggfinalfn => 'numeric_stddev_pop', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'numeric_accum', aggminvtransfn => 'numeric_accum_inv', aggmfinalfn => 'numeric_stddev_pop', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'stddev_pop_p_numeric' }, # stddev_samp +{ aggfnoid => 'stddev_samp_p_int8', aggtransfn => 'int8_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'stddev_samp_p_int8' }, { aggfnoid => 'stddev_samp(int8)', aggtransfn => 'int8_accum', aggfinalfn => 'numeric_stddev_samp', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'int8_accum', aggminvtransfn => 'int8_accum_inv', aggmfinalfn => 'numeric_stddev_samp', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'stddev_samp_p_int8' }, +{ aggfnoid => 'stddev_samp_p_int4', aggtransfn => 'int4_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'stddev_samp_p_int4' }, { aggfnoid => 'stddev_samp(int4)', aggtransfn => 'int4_accum', aggfinalfn => 'numeric_poly_stddev_samp', aggcombinefn => 'numeric_poly_combine', @@ -391,7 +650,14 @@ aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int4_accum', aggminvtransfn => 'int4_accum_inv', aggmfinalfn => 'numeric_poly_stddev_samp', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'stddev_samp_p_int4' }, +{ aggfnoid => 'stddev_samp_p_int2', aggtransfn => 'int2_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'stddev_samp_p_int2' }, { aggfnoid => 'stddev_samp(int2)', aggtransfn => 'int2_accum', aggfinalfn => 'numeric_poly_stddev_samp', aggcombinefn => 'numeric_poly_combine', @@ -399,29 +665,58 @@ aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int2_accum', aggminvtransfn => 'int2_accum_inv', aggmfinalfn => 'numeric_poly_stddev_samp', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'stddev_samp_p_int2' }, +{ aggfnoid => 'stddev_samp_p_float4', aggtransfn => 'float4_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'stddev_samp_p_float4' }, { aggfnoid => 'stddev_samp(float4)', aggtransfn => 'float4_accum', aggfinalfn => 'float8_stddev_samp', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'stddev_samp_p_float4' }, +{ aggfnoid => 'stddev_samp_p_float8', aggtransfn => 'float8_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'stddev_samp_p_float8' }, { aggfnoid => 'stddev_samp(float8)', aggtransfn => 'float8_accum', aggfinalfn => 'float8_stddev_samp', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'stddev_samp_p_float8' }, +{ aggfnoid => 'stddev_samp_p_numeric', aggtransfn => 'numeric_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'stddev_samp_p_numeric' }, { aggfnoid => 'stddev_samp(numeric)', aggtransfn => 'numeric_accum', aggfinalfn => 'numeric_stddev_samp', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'numeric_accum', aggminvtransfn => 'numeric_accum_inv', aggmfinalfn => 'numeric_stddev_samp', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'stddev_samp_p_numeric' }, # stddev: historical Postgres syntax for stddev_samp +{ aggfnoid => 'stddev_p_int8', aggtransfn => 'int8_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'stddev_p_int8' }, { aggfnoid => 'stddev(int8)', aggtransfn => 'int8_accum', aggfinalfn => 'numeric_stddev_samp', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'int8_accum', aggminvtransfn => 'int8_accum_inv', aggmfinalfn => 'numeric_stddev_samp', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'stddev_p_int8' }, +{ aggfnoid => 'stddev_p_int4', aggtransfn => 'int4_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'stddev_p_int4' }, { aggfnoid => 'stddev(int4)', aggtransfn => 'int4_accum', aggfinalfn => 'numeric_poly_stddev_samp', aggcombinefn => 'numeric_poly_combine', @@ -429,7 +724,14 @@ aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int4_accum', aggminvtransfn => 'int4_accum_inv', aggmfinalfn => 'numeric_poly_stddev_samp', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'stddev_p_int4' }, +{ aggfnoid => 'stddev_p_int2', aggtransfn => 'int2_accum', + aggfinalfn => 'numeric_poly_serialize', aggcombinefn => 'numeric_poly_combine', + aggserialfn => 'numeric_poly_serialize', + aggdeserialfn => 'numeric_poly_deserialize', aggtranstype => 'internal', + aggtransspace => '48', + aggpartialfn => 'stddev_p_int2' }, { aggfnoid => 'stddev(int2)', aggtransfn => 'int2_accum', aggfinalfn => 'numeric_poly_stddev_samp', aggcombinefn => 'numeric_poly_combine', @@ -437,138 +739,253 @@ aggdeserialfn => 'numeric_poly_deserialize', aggmtransfn => 'int2_accum', aggminvtransfn => 'int2_accum_inv', aggmfinalfn => 'numeric_poly_stddev_samp', aggtranstype => 'internal', aggtransspace => '48', - aggmtranstype => 'internal', aggmtransspace => '48' }, + aggmtranstype => 'internal', aggmtransspace => '48', + aggpartialfn => 'stddev_p_int2' }, +{ aggfnoid => 'stddev_p_float4', aggtransfn => 'float4_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'stddev_p_float4' }, { aggfnoid => 'stddev(float4)', aggtransfn => 'float4_accum', aggfinalfn => 'float8_stddev_samp', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'stddev_p_float4' }, +{ aggfnoid => 'stddev_p_float8', aggtransfn => 'float8_accum', + aggcombinefn => 'float8_combine', aggtranstype => '_float8', + agginitval => '{0,0,0}', + aggpartialfn => 'stddev_p_float8' }, { aggfnoid => 'stddev(float8)', aggtransfn => 'float8_accum', aggfinalfn => 'float8_stddev_samp', aggcombinefn => 'float8_combine', - aggtranstype => '_float8', agginitval => '{0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0}', + aggpartialfn => 'stddev_p_float8' }, +{ aggfnoid => 'stddev_p_numeric', aggtransfn => 'numeric_accum', + aggfinalfn => 'numeric_serialize', aggcombinefn => 'numeric_combine', + aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', + aggtranstype => 'internal', aggtransspace => '128', + aggpartialfn => 'stddev_p_numeric' }, { aggfnoid => 'stddev(numeric)', aggtransfn => 'numeric_accum', aggfinalfn => 'numeric_stddev_samp', aggcombinefn => 'numeric_combine', aggserialfn => 'numeric_serialize', aggdeserialfn => 'numeric_deserialize', aggmtransfn => 'numeric_accum', aggminvtransfn => 'numeric_accum_inv', aggmfinalfn => 'numeric_stddev_samp', aggtranstype => 'internal', aggtransspace => '128', aggmtranstype => 'internal', - aggmtransspace => '128' }, + aggmtransspace => '128', + aggpartialfn => 'stddev_p_numeric' }, # SQL2003 binary regression aggregates { aggfnoid => 'regr_count', aggtransfn => 'int8inc_float8_float8', - aggcombinefn => 'int8pl', aggtranstype => 'int8', agginitval => '0' }, + aggcombinefn => 'int8pl', aggtranstype => 'int8', agginitval => '0', + aggpartialfn => 'regr_count' }, +{ aggfnoid => 'regr_sxx_p', aggtransfn => 'float8_regr_accum', + aggcombinefn => 'float8_regr_combine', aggtranstype => '_float8', + agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_sxx_p' }, { aggfnoid => 'regr_sxx', aggtransfn => 'float8_regr_accum', aggfinalfn => 'float8_regr_sxx', aggcombinefn => 'float8_regr_combine', - aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_sxx_p' }, +{ aggfnoid => 'regr_syy_p', aggtransfn => 'float8_regr_accum', + aggcombinefn => 'float8_regr_combine', aggtranstype => '_float8', + agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_syy_p' }, { aggfnoid => 'regr_syy', aggtransfn => 'float8_regr_accum', aggfinalfn => 'float8_regr_syy', aggcombinefn => 'float8_regr_combine', - aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_syy_p' }, +{ aggfnoid => 'regr_sxy_p', aggtransfn => 'float8_regr_accum', + aggcombinefn => 'float8_regr_combine', aggtranstype => '_float8', + agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_sxy_p' }, { aggfnoid => 'regr_sxy', aggtransfn => 'float8_regr_accum', aggfinalfn => 'float8_regr_sxy', aggcombinefn => 'float8_regr_combine', - aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_sxy_p' }, +{ aggfnoid => 'regr_avgx_p', aggtransfn => 'float8_regr_accum', + aggcombinefn => 'float8_regr_combine', aggtranstype => '_float8', + agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_avgx_p' }, { aggfnoid => 'regr_avgx', aggtransfn => 'float8_regr_accum', aggfinalfn => 'float8_regr_avgx', aggcombinefn => 'float8_regr_combine', - aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_avgx_p' }, +{ aggfnoid => 'regr_avgy_p', aggtransfn => 'float8_regr_accum', + aggcombinefn => 'float8_regr_combine', aggtranstype => '_float8', + agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_avgy_p' }, { aggfnoid => 'regr_avgy', aggtransfn => 'float8_regr_accum', aggfinalfn => 'float8_regr_avgy', aggcombinefn => 'float8_regr_combine', - aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_avgy_p' }, +{ aggfnoid => 'regr_r2_p', aggtransfn => 'float8_regr_accum', + aggcombinefn => 'float8_regr_combine', aggtranstype => '_float8', + agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_r2_p' }, { aggfnoid => 'regr_r2', aggtransfn => 'float8_regr_accum', aggfinalfn => 'float8_regr_r2', aggcombinefn => 'float8_regr_combine', - aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_r2_p' }, +{ aggfnoid => 'regr_slope_p', aggtransfn => 'float8_regr_accum', + aggcombinefn => 'float8_regr_combine', aggtranstype => '_float8', + agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_slope_p' }, { aggfnoid => 'regr_slope', aggtransfn => 'float8_regr_accum', aggfinalfn => 'float8_regr_slope', aggcombinefn => 'float8_regr_combine', - aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_slope_p' }, +{ aggfnoid => 'regr_intercept_p', aggtransfn => 'float8_regr_accum', + aggcombinefn => 'float8_regr_combine', aggtranstype => '_float8', + agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_intercept_p' }, { aggfnoid => 'regr_intercept', aggtransfn => 'float8_regr_accum', aggfinalfn => 'float8_regr_intercept', aggcombinefn => 'float8_regr_combine', - aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'regr_intercept_p' }, +{ aggfnoid => 'covar_pop_p', aggtransfn => 'float8_regr_accum', + aggcombinefn => 'float8_regr_combine', aggtranstype => '_float8', + agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'covar_pop_p' }, { aggfnoid => 'covar_pop', aggtransfn => 'float8_regr_accum', aggfinalfn => 'float8_covar_pop', aggcombinefn => 'float8_regr_combine', - aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'covar_pop_p' }, +{ aggfnoid => 'covar_samp_p', aggtransfn => 'float8_regr_accum', + aggcombinefn => 'float8_regr_combine', aggtranstype => '_float8', + agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'covar_samp_p' }, { aggfnoid => 'covar_samp', aggtransfn => 'float8_regr_accum', aggfinalfn => 'float8_covar_samp', aggcombinefn => 'float8_regr_combine', - aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'covar_samp_p' }, +{ aggfnoid => 'corr_p', aggtransfn => 'float8_regr_accum', + aggcombinefn => 'float8_regr_combine', aggtranstype => '_float8', + agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'corr_p' }, { aggfnoid => 'corr', aggtransfn => 'float8_regr_accum', aggfinalfn => 'float8_corr', aggcombinefn => 'float8_regr_combine', - aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}' }, + aggtranstype => '_float8', agginitval => '{0,0,0,0,0,0}', + aggpartialfn => 'corr_p' }, # boolean-and and boolean-or { aggfnoid => 'bool_and', aggtransfn => 'booland_statefunc', aggcombinefn => 'booland_statefunc', aggmtransfn => 'bool_accum', aggminvtransfn => 'bool_accum_inv', aggmfinalfn => 'bool_alltrue', aggsortop => '<(bool,bool)', aggtranstype => 'bool', - aggmtranstype => 'internal', aggmtransspace => '16' }, + aggmtranstype => 'internal', aggmtransspace => '16', + aggpartialfn => 'bool_and' }, { aggfnoid => 'bool_or', aggtransfn => 'boolor_statefunc', aggcombinefn => 'boolor_statefunc', aggmtransfn => 'bool_accum', aggminvtransfn => 'bool_accum_inv', aggmfinalfn => 'bool_anytrue', aggsortop => '>(bool,bool)', aggtranstype => 'bool', - aggmtranstype => 'internal', aggmtransspace => '16' }, + aggmtranstype => 'internal', aggmtransspace => '16', + aggpartialfn => 'bool_or' }, { aggfnoid => 'every', aggtransfn => 'booland_statefunc', aggcombinefn => 'booland_statefunc', aggmtransfn => 'bool_accum', aggminvtransfn => 'bool_accum_inv', aggmfinalfn => 'bool_alltrue', aggsortop => '<(bool,bool)', aggtranstype => 'bool', - aggmtranstype => 'internal', aggmtransspace => '16' }, + aggmtranstype => 'internal', aggmtransspace => '16', + aggpartialfn => 'every' }, # bitwise integer { aggfnoid => 'bit_and(int2)', aggtransfn => 'int2and', - aggcombinefn => 'int2and', aggtranstype => 'int2' }, + aggcombinefn => 'int2and', aggtranstype => 'int2', + aggpartialfn => 'bit_and(int2)' }, { aggfnoid => 'bit_or(int2)', aggtransfn => 'int2or', aggcombinefn => 'int2or', - aggtranstype => 'int2' }, + aggtranstype => 'int2', + aggpartialfn => 'bit_or(int2)' }, { aggfnoid => 'bit_xor(int2)', aggtransfn => 'int2xor', - aggcombinefn => 'int2xor', aggtranstype => 'int2' }, + aggcombinefn => 'int2xor', aggtranstype => 'int2', + aggpartialfn => 'bit_xor(int2)' }, { aggfnoid => 'bit_and(int4)', aggtransfn => 'int4and', - aggcombinefn => 'int4and', aggtranstype => 'int4' }, + aggcombinefn => 'int4and', aggtranstype => 'int4', + aggpartialfn => 'bit_and(int4)' }, { aggfnoid => 'bit_or(int4)', aggtransfn => 'int4or', aggcombinefn => 'int4or', - aggtranstype => 'int4' }, + aggtranstype => 'int4', + aggpartialfn => 'bit_or(int4)' }, { aggfnoid => 'bit_xor(int4)', aggtransfn => 'int4xor', - aggcombinefn => 'int4xor', aggtranstype => 'int4' }, + aggcombinefn => 'int4xor', aggtranstype => 'int4', + aggpartialfn => 'bit_xor(int4)' }, { aggfnoid => 'bit_and(int8)', aggtransfn => 'int8and', - aggcombinefn => 'int8and', aggtranstype => 'int8' }, + aggcombinefn => 'int8and', aggtranstype => 'int8', + aggpartialfn => 'bit_and(int8)' }, { aggfnoid => 'bit_or(int8)', aggtransfn => 'int8or', aggcombinefn => 'int8or', - aggtranstype => 'int8' }, + aggtranstype => 'int8', + aggpartialfn => 'bit_or(int8)' }, { aggfnoid => 'bit_xor(int8)', aggtransfn => 'int8xor', - aggcombinefn => 'int8xor', aggtranstype => 'int8' }, + aggcombinefn => 'int8xor', aggtranstype => 'int8', + aggpartialfn => 'bit_xor(int8)' }, { aggfnoid => 'bit_and(bit)', aggtransfn => 'bitand', aggcombinefn => 'bitand', - aggtranstype => 'bit' }, + aggtranstype => 'bit', + aggpartialfn => 'bit_and(bit)' }, { aggfnoid => 'bit_or(bit)', aggtransfn => 'bitor', aggcombinefn => 'bitor', - aggtranstype => 'bit' }, + aggtranstype => 'bit', + aggpartialfn => 'bit_or(bit)' }, { aggfnoid => 'bit_xor(bit)', aggtransfn => 'bitxor', aggcombinefn => 'bitxor', - aggtranstype => 'bit' }, + aggtranstype => 'bit', + aggpartialfn => 'bit_xor(bit)' }, # xml { aggfnoid => 'xmlagg', aggtransfn => 'xmlconcat2', aggtranstype => 'xml' }, # array +{ aggfnoid => 'array_agg_p_anynonarray', aggtransfn => 'array_agg_transfn', + aggfinalfn => 'array_agg_serialize', aggcombinefn => 'array_agg_combine', + aggserialfn => 'array_agg_serialize', aggdeserialfn => 'array_agg_deserialize', + aggtranstype => 'internal', + aggpartialfn => 'array_agg_p_anynonarray' }, { aggfnoid => 'array_agg(anynonarray)', aggtransfn => 'array_agg_transfn', aggfinalfn => 'array_agg_finalfn', aggcombinefn => 'array_agg_combine', aggserialfn => 'array_agg_serialize', aggdeserialfn => 'array_agg_deserialize', aggfinalextra => 't', - aggtranstype => 'internal' }, + aggtranstype => 'internal', + aggpartialfn => 'array_agg_p_anynonarray' }, +{ aggfnoid => 'array_agg_p_anyarray', aggtransfn => 'array_agg_array_transfn', + aggfinalfn => 'array_agg_array_serialize', + aggcombinefn => 'array_agg_array_combine', + aggserialfn => 'array_agg_array_serialize', + aggdeserialfn => 'array_agg_array_deserialize', aggtranstype => 'internal', + aggpartialfn => 'array_agg_p_anyarray' }, { aggfnoid => 'array_agg(anyarray)', aggtransfn => 'array_agg_array_transfn', aggfinalfn => 'array_agg_array_finalfn', aggcombinefn => 'array_agg_array_combine', aggserialfn => 'array_agg_array_serialize', aggdeserialfn => 'array_agg_array_deserialize', aggfinalextra => 't', - aggtranstype => 'internal' }, + aggtranstype => 'internal', + aggpartialfn => 'array_agg_p_anyarray' }, # text +{ aggfnoid => 'string_agg_p_text_text', aggtransfn => 'string_agg_transfn', + aggfinalfn => 'string_agg_serialize', aggcombinefn => 'string_agg_combine', + aggserialfn => 'string_agg_serialize', + aggdeserialfn => 'string_agg_deserialize', aggtranstype => 'internal', + aggpartialfn => 'string_agg_p_text_text' }, { aggfnoid => 'string_agg(text,text)', aggtransfn => 'string_agg_transfn', aggfinalfn => 'string_agg_finalfn', aggcombinefn => 'string_agg_combine', aggserialfn => 'string_agg_serialize', - aggdeserialfn => 'string_agg_deserialize', aggtranstype => 'internal' }, + aggdeserialfn => 'string_agg_deserialize', aggtranstype => 'internal', + aggpartialfn => 'string_agg_p_text_text' }, # bytea +{ aggfnoid => 'string_agg_p_bytea_bytea', + aggtransfn => 'bytea_string_agg_transfn', aggfinalfn => 'string_agg_serialize', + aggcombinefn => 'string_agg_combine', aggserialfn => 'string_agg_serialize', + aggdeserialfn => 'string_agg_deserialize', aggtranstype => 'internal', + aggpartialfn => 'string_agg_p_bytea_bytea' }, { aggfnoid => 'string_agg(bytea,bytea)', aggtransfn => 'bytea_string_agg_transfn', aggfinalfn => 'bytea_string_agg_finalfn', aggcombinefn => 'string_agg_combine', aggserialfn => 'string_agg_serialize', - aggdeserialfn => 'string_agg_deserialize', aggtranstype => 'internal' }, + aggdeserialfn => 'string_agg_deserialize', aggtranstype => 'internal', + aggpartialfn => 'string_agg_p_bytea_bytea' }, # range { aggfnoid => 'range_intersect_agg(anyrange)', aggtransfn => 'range_intersect_agg_transfn', - aggcombinefn => 'range_intersect_agg_transfn', aggtranstype => 'anyrange' }, + aggcombinefn => 'range_intersect_agg_transfn', aggtranstype => 'anyrange', + aggpartialfn => 'range_intersect_agg(anyrange)' }, { aggfnoid => 'range_intersect_agg(anymultirange)', aggtransfn => 'multirange_intersect_agg_transfn', aggcombinefn => 'multirange_intersect_agg_transfn', - aggtranstype => 'anymultirange' }, + aggtranstype => 'anymultirange', + aggpartialfn => 'range_intersect_agg(anymultirange)' }, { aggfnoid => 'range_agg(anyrange)', aggtransfn => 'range_agg_transfn', aggfinalfn => 'range_agg_finalfn', aggfinalextra => 't', aggtranstype => 'internal' }, @@ -658,6 +1075,7 @@ # any_value { aggfnoid => 'any_value(anyelement)', aggtransfn => 'any_value_transfn', - aggcombinefn => 'any_value_transfn', aggtranstype => 'anyelement' }, + aggcombinefn => 'any_value_transfn', aggtranstype => 'anyelement', + aggpartialfn => 'any_value(anyelement)' }, ] diff --git a/src/include/catalog/pg_aggregate.h b/src/include/catalog/pg_aggregate.h index 3112881193..b8fb0372b3 100644 --- a/src/include/catalog/pg_aggregate.h +++ b/src/include/catalog/pg_aggregate.h @@ -55,6 +55,9 @@ CATALOG(pg_aggregate,2600,AggregateRelationId) /* function to convert bytea to transtype (0 if none) */ regproc aggdeserialfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc); + /* special aggregate function for partial aggregation (0 if none) */ + regproc aggpartialfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc); + /* forward function for moving-aggregate mode (0 if none) */ regproc aggmtransfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc); @@ -164,6 +167,7 @@ extern ObjectAddress AggregateCreate(const char *aggName, List *aggmtransfnName, List *aggminvtransfnName, List *aggmfinalfnName, + List *aggpartialfnName, bool finalfnExtraArgs, bool mfinalfnExtraArgs, char finalfnModify, diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat index 6996073989..1c4087ad8c 100644 --- a/src/include/catalog/pg_proc.dat +++ b/src/include/catalog/pg_proc.dat @@ -1658,6 +1658,10 @@ { oid => '2334', descr => 'aggregate final function', proname => 'array_agg_finalfn', proisstrict => 'f', prorettype => 'anyarray', proargtypes => 'internal anynonarray', prosrc => 'array_agg_finalfn' }, +{ oid => '6', descr => 'partial aggregation for array_agg(anynonarray)', + proname => 'array_agg_p_anynonarray', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'anynonarray', + prosrc => 'aggregate_dummy' }, { oid => '2335', descr => 'concatenate aggregate input into an array', proname => 'array_agg', prokind => 'a', proisstrict => 'f', prorettype => 'anyarray', proargtypes => 'anynonarray', @@ -1680,6 +1684,9 @@ proname => 'array_agg_array_finalfn', proisstrict => 'f', prorettype => 'anyarray', proargtypes => 'internal anyarray', prosrc => 'array_agg_array_finalfn' }, +{ oid => '7', descr => 'partial aggregation for array_agg(anyarray)', + proname => 'array_agg_p_anyarray', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'anyarray', prosrc => 'aggregate_dummy' }, { oid => '4053', descr => 'concatenate aggregate input into an array', proname => 'array_agg', prokind => 'a', proisstrict => 'f', prorettype => 'anyarray', proargtypes => 'anyarray', @@ -4989,6 +4996,10 @@ { oid => '3536', descr => 'aggregate final function', proname => 'string_agg_finalfn', proisstrict => 'f', prorettype => 'text', proargtypes => 'internal', prosrc => 'string_agg_finalfn' }, +{ oid => '8', descr => 'partial aggregation for string_agg(text, text)', + proname => 'string_agg_p_text_text', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'text text', + prosrc => 'aggregate_dummy' }, { oid => '3538', descr => 'concatenate aggregate input into a string', proname => 'string_agg', prokind => 'a', proisstrict => 'f', prorettype => 'text', proargtypes => 'text text', @@ -5001,6 +5012,10 @@ proname => 'bytea_string_agg_finalfn', proisstrict => 'f', prorettype => 'bytea', proargtypes => 'internal', prosrc => 'bytea_string_agg_finalfn' }, +{ oid => '9', descr => 'partial aggregation for string_agg(bytea, bytea)', + proname => 'string_agg_p_bytea_bytea', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'bytea bytea', + prosrc => 'aggregate_dummy' }, { oid => '3545', descr => 'concatenate aggregate input into a bytea', proname => 'string_agg', prokind => 'a', proisstrict => 'f', prorettype => 'bytea', proargtypes => 'bytea bytea', @@ -6579,36 +6594,61 @@ # Aggregates (moved here from pg_aggregate for 7.3) +{ oid => '111', descr => 'partial aggregation for avg(int8)', + proname => 'avg_p_int8', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, { oid => '2100', descr => 'the average (arithmetic mean) as numeric of all bigint values', proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, +{ oid => '226', descr => 'partial aggregation for avg(int4)', + proname => 'avg_p_int4', prokind => 'a', proisstrict => 'f', + prorettype => '_int8', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, { oid => '2101', descr => 'the average (arithmetic mean) as numeric of all integer values', proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, +{ oid => '388', descr => 'partial aggregation for avg(int2)', + proname => 'avg_p_int2', prokind => 'a', proisstrict => 'f', + prorettype => '_int8', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, { oid => '2102', descr => 'the average (arithmetic mean) as numeric of all smallint values', proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, +{ oid => '389', descr => 'partial aggregation for avg(numeric)', + proname => 'avg_p_numeric', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, { oid => '2103', descr => 'the average (arithmetic mean) as numeric of all numeric values', proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, +{ oid => '560', descr => 'partial aggregation for avg(float4)', + proname => 'avg_p_float4', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, { oid => '2104', descr => 'the average (arithmetic mean) as float8 of all float4 values', proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, +{ oid => '561', descr => 'partial aggregation for avg(float8)', + proname => 'avg_p_float8', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, { oid => '2105', descr => 'the average (arithmetic mean) as float8 of all float8 values', proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, +{ oid => '562', descr => 'partial aggregation for avg(interval)', + proname => 'avg_p_interval', prokind => 'a', proisstrict => 'f', + prorettype => '_interval', proargtypes => 'interval', + prosrc => 'aggregate_dummy' }, { oid => '2106', descr => 'the average (arithmetic mean) as interval of all interval values', proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'interval', proargtypes => 'interval', prosrc => 'aggregate_dummy' }, +{ oid => '563', descr => 'partial aggregation for sum(int8)', + proname => 'sum_p_int8', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, { oid => '2107', descr => 'sum as numeric across all bigint input values', proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, @@ -6631,6 +6671,9 @@ proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'interval', proargtypes => 'interval', prosrc => 'aggregate_dummy' }, +{ oid => '564', descr => 'partial aggregation for sum(numeric)', + proname => 'sum_p_numeric', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, { oid => '2114', descr => 'sum as numeric across all numeric input values', proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, @@ -6789,153 +6832,261 @@ proname => 'int8inc_support', prorettype => 'internal', proargtypes => 'internal', prosrc => 'int8inc_support' }, +{ oid => '565', descr => 'partial aggregation for var_pop(int8)', + proname => 'var_pop_p_int8', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, { oid => '2718', descr => 'population variance of bigint input values (square of the population standard deviation)', proname => 'var_pop', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, +{ oid => '566', descr => 'partial aggregation for var_pop(int4)', + proname => 'var_pop_p_int4', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, { oid => '2719', descr => 'population variance of integer input values (square of the population standard deviation)', proname => 'var_pop', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, +{ oid => '567', descr => 'partial aggregation for var_pop(int2)', + proname => 'var_pop_p_int2', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, { oid => '2720', descr => 'population variance of smallint input values (square of the population standard deviation)', proname => 'var_pop', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, +{ oid => '568', descr => 'partial aggregation for var_pop(float4)', + proname => 'var_pop_p_float4', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, { oid => '2721', descr => 'population variance of float4 input values (square of the population standard deviation)', proname => 'var_pop', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, +{ oid => '569', descr => 'partial aggregation for var_pop(float8)', + proname => 'var_pop_p_float8', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, { oid => '2722', descr => 'population variance of float8 input values (square of the population standard deviation)', proname => 'var_pop', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, +{ oid => '570', descr => 'partial aggregation for var_pop(numeric)', + proname => 'var_pop_p_numeric', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, { oid => '2723', descr => 'population variance of numeric input values (square of the population standard deviation)', proname => 'var_pop', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, +{ oid => '571', descr => 'partial aggregation for var_samp(int8)', + proname => 'var_samp_p_int8', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, { oid => '2641', descr => 'sample variance of bigint input values (square of the sample standard deviation)', proname => 'var_samp', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, +{ oid => '572', descr => 'partial aggregation for var_samp(int4)', + proname => 'var_samp_p_int4', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, { oid => '2642', descr => 'sample variance of integer input values (square of the sample standard deviation)', proname => 'var_samp', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, +{ oid => '573', descr => 'partial aggregation for var_samp(int2)', + proname => 'var_samp_p_int2', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, { oid => '2643', descr => 'sample variance of smallint input values (square of the sample standard deviation)', proname => 'var_samp', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, +{ oid => '574', descr => 'partial aggregation for var_samp(float4)', + proname => 'var_samp_p_float4', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, { oid => '2644', descr => 'sample variance of float4 input values (square of the sample standard deviation)', proname => 'var_samp', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, +{ oid => '575', descr => 'partial aggregation for var_samp(float8)', + proname => 'var_samp_p_float8', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, { oid => '2645', descr => 'sample variance of float8 input values (square of the sample standard deviation)', proname => 'var_samp', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, +{ oid => '576', descr => 'partial aggregation for var_samp(numeric)', + proname => 'var_samp_p_numeric', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, { oid => '2646', descr => 'sample variance of numeric input values (square of the sample standard deviation)', proname => 'var_samp', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, +{ oid => '577', descr => 'partial aggregation for variance(int8)', + proname => 'variance_p_int8', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, { oid => '2148', descr => 'historical alias for var_samp', proname => 'variance', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, +{ oid => '578', descr => 'partial aggregation for variance(int4)', + proname => 'variance_p_int4', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, { oid => '2149', descr => 'historical alias for var_samp', proname => 'variance', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, +{ oid => '579', descr => 'partial aggregation for variance(int2)', + proname => 'variance_p_int2', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, { oid => '2150', descr => 'historical alias for var_samp', proname => 'variance', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, +{ oid => '580', descr => 'partial aggregation for variance(float4)', + proname => 'variance_p_float4', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, { oid => '2151', descr => 'historical alias for var_samp', proname => 'variance', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, +{ oid => '581', descr => 'partial aggregation for variance(float8)', + proname => 'variance_p_float8', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, { oid => '2152', descr => 'historical alias for var_samp', proname => 'variance', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, +{ oid => '582', descr => 'partial aggregation for variance(numeric)', + proname => 'variance_p_numeric', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, { oid => '2153', descr => 'historical alias for var_samp', proname => 'variance', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, +{ oid => '583', descr => 'partial aggregation for stddev_pop(int8)', + proname => 'stddev_pop_p_int8', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, { oid => '2724', descr => 'population standard deviation of bigint input values', proname => 'stddev_pop', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, +{ oid => '703', descr => 'partial aggregation for stddev_pop(int4)', + proname => 'stddev_pop_p_int4', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, { oid => '2725', descr => 'population standard deviation of integer input values', proname => 'stddev_pop', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, +{ oid => '726', descr => 'partial aggregation for stddev_pop(int2)', + proname => 'stddev_pop_p_int2', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, { oid => '2726', descr => 'population standard deviation of smallint input values', proname => 'stddev_pop', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, +{ oid => '786', descr => 'partial aggregation for stddev_pop(float4)', + proname => 'stddev_pop_p_float4', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, { oid => '2727', descr => 'population standard deviation of float4 input values', proname => 'stddev_pop', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, +{ oid => '787', descr => 'partial aggregation for stddev_pop(float8)', + proname => 'stddev_pop_p_float8', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, { oid => '2728', descr => 'population standard deviation of float8 input values', proname => 'stddev_pop', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, +{ oid => '788', descr => 'partial aggregation for stddev_pop(numeric)', + proname => 'stddev_pop_p_numeric', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, { oid => '2729', descr => 'population standard deviation of numeric input values', proname => 'stddev_pop', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, +{ oid => '789', descr => 'partial aggregation for stddev_samp(int8)', + proname => 'stddev_samp_p_int8', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, { oid => '2712', descr => 'sample standard deviation of bigint input values', proname => 'stddev_samp', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, +{ oid => '811', descr => 'partial aggregation for stddev_samp(int4)', + proname => 'stddev_samp_p_int4', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, { oid => '2713', descr => 'sample standard deviation of integer input values', proname => 'stddev_samp', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, +{ oid => '812', descr => 'partial aggregation for stddev_samp(int2)', + proname => 'stddev_samp_p_int2', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, { oid => '2714', descr => 'sample standard deviation of smallint input values', proname => 'stddev_samp', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, +{ oid => '813', descr => 'partial aggregation for stddev_samp(float4)', + proname => 'stddev_samp_p_float4', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, { oid => '2715', descr => 'sample standard deviation of float4 input values', proname => 'stddev_samp', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, +{ oid => '814', descr => 'partial aggregation for stddev_samp(float8)', + proname => 'stddev_samp_p_float8', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, { oid => '2716', descr => 'sample standard deviation of float8 input values', proname => 'stddev_samp', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, +{ oid => '815', descr => 'partial aggregation for stddev_samp(numeric)', + proname => 'stddev_samp_p_numeric', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, { oid => '2717', descr => 'sample standard deviation of numeric input values', proname => 'stddev_samp', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, +{ oid => '816', descr => 'partial aggregation for stddev(int8)', + proname => 'stddev_p_int8', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, { oid => '2154', descr => 'historical alias for stddev_samp', proname => 'stddev', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int8', prosrc => 'aggregate_dummy' }, +{ oid => '970', descr => 'partial aggregation for stddev(int4)', + proname => 'stddev_p_int4', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, { oid => '2155', descr => 'historical alias for stddev_samp', proname => 'stddev', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int4', prosrc => 'aggregate_dummy' }, +{ oid => '1382', descr => 'partial aggregation for stddev(int2)', + proname => 'stddev_p_int2', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, { oid => '2156', descr => 'historical alias for stddev_samp', proname => 'stddev', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'int2', prosrc => 'aggregate_dummy' }, +{ oid => '1524', descr => 'partial aggregation for stddev(float4)', + proname => 'stddev_p_float4', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, { oid => '2157', descr => 'historical alias for stddev_samp', proname => 'stddev', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float4', prosrc => 'aggregate_dummy' }, +{ oid => '1533', descr => 'partial aggregation for stddev(float8)', + proname => 'stddev_p_float8', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, { oid => '2158', descr => 'historical alias for stddev_samp', proname => 'stddev', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8', prosrc => 'aggregate_dummy' }, +{ oid => '1566', descr => 'partial aggregation for stddev(numeric)', + proname => 'stddev_p_numeric', prokind => 'a', proisstrict => 'f', + prorettype => 'bytea', proargtypes => 'numeric', prosrc => 'aggregate_dummy' }, { oid => '2159', descr => 'historical alias for stddev_samp', proname => 'stddev', prokind => 'a', proisstrict => 'f', prorettype => 'numeric', proargtypes => 'numeric', @@ -6946,52 +7097,97 @@ proname => 'regr_count', prokind => 'a', proisstrict => 'f', prorettype => 'int8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, +{ oid => '1568', descr => 'partial aggregation for regr_sxx(float8, float8)', + proname => 'regr_sxx_p', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8 float8', + prosrc => 'aggregate_dummy' }, { oid => '2819', descr => 'sum of squares of the independent variable (sum(X^2) - sum(X)^2/N)', proname => 'regr_sxx', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, +{ oid => '2173', descr => 'partial aggregation for regr_syy(float8, float8)', + proname => 'regr_syy_p', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8 float8', + prosrc => 'aggregate_dummy' }, { oid => '2820', descr => 'sum of squares of the dependent variable (sum(Y^2) - sum(Y)^2/N)', proname => 'regr_syy', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, +{ oid => '3813', descr => 'partial aggregation for regr_sxy(float8, float8)', + proname => 'regr_sxy_p', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8 float8', + prosrc => 'aggregate_dummy' }, { oid => '2821', descr => 'sum of products of independent times dependent variable (sum(X*Y) - sum(X) * sum(Y)/N)', proname => 'regr_sxy', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, +{ oid => '3814', descr => 'partial aggregation for regr_avgx(float8, float8)', + proname => 'regr_avgx_p', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8 float8', + prosrc => 'aggregate_dummy' }, { oid => '2822', descr => 'average of the independent variable (sum(X)/N)', proname => 'regr_avgx', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, +{ oid => '4551', descr => 'partial aggregation for regr_avgy(float8, float8)', + proname => 'regr_avgy_p', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8 float8', + prosrc => 'aggregate_dummy' }, { oid => '2823', descr => 'average of the dependent variable (sum(Y)/N)', proname => 'regr_avgy', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, +{ oid => '4552', descr => 'partial aggregation for regr_r2(float8, float8)', + proname => 'regr_r2_p', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8 float8', + prosrc => 'aggregate_dummy' }, { oid => '2824', descr => 'square of the correlation coefficient', proname => 'regr_r2', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, +{ oid => '4553', descr => 'partial aggregation for regr_slope(float8, float8)', + proname => 'regr_slope_p', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8 float8', + prosrc => 'aggregate_dummy' }, { oid => '2825', descr => 'slope of the least-squares-fit linear equation determined by the (X, Y) pairs', proname => 'regr_slope', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, +{ oid => '4554', + descr => 'partial aggregation for regr_intercept(float8, float8)', + proname => 'regr_intercept_p', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8 float8', + prosrc => 'aggregate_dummy' }, { oid => '2826', descr => 'y-intercept of the least-squares-fit linear equation determined by the (X, Y) pairs', proname => 'regr_intercept', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, +{ oid => '4555', descr => 'partial aggregation for covar_pop(float8, float8)', + proname => 'covar_pop_p', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8 float8', + prosrc => 'aggregate_dummy' }, { oid => '2827', descr => 'population covariance', proname => 'covar_pop', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, +{ oid => '4556', descr => 'partial aggregation for covar_samp(float8, float8)', + proname => 'covar_samp_p', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8 float8', + prosrc => 'aggregate_dummy' }, { oid => '2828', descr => 'sample covariance', proname => 'covar_samp', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, +{ oid => '4557', descr => 'partial aggregation for corr(float8, float8)', + proname => 'corr_p', prokind => 'a', proisstrict => 'f', + prorettype => '_float8', proargtypes => 'float8 float8', + prosrc => 'aggregate_dummy' }, { oid => '2829', descr => 'correlation coefficient', proname => 'corr', prokind => 'a', proisstrict => 'f', prorettype => 'float8', proargtypes => 'float8 float8', prosrc => 'aggregate_dummy' }, diff --git a/src/include/nodes/pathnodes.h b/src/include/nodes/pathnodes.h index c17b53f7ad..443de6fb24 100644 --- a/src/include/nodes/pathnodes.h +++ b/src/include/nodes/pathnodes.h @@ -3244,6 +3244,8 @@ typedef struct Node *havingQual; List *targetList; PartitionwiseAggregateType patype; + List *groupClausePartial; + PathTarget *partial_target; } GroupPathExtraData; /* diff --git a/src/test/regress/expected/create_aggregate.out b/src/test/regress/expected/create_aggregate.out index dcf6909423..73e4cfde0f 100644 --- a/src/test/regress/expected/create_aggregate.out +++ b/src/test/regress/expected/create_aggregate.out @@ -322,3 +322,187 @@ WARNING: aggregate attribute "Finalfunc_extra" not recognized WARNING: aggregate attribute "Finalfunc_modify" not recognized WARNING: aggregate attribute "Parallel" not recognized ERROR: aggregate stype must be specified +-- invalid: finalfunc is specified and stype is not internal +-- even though aggpartialfunc is equal to the aggregate name +CREATE AGGREGATE udf_avg_int4_invalid(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + initcond = '{0,0}', + aggpartialfunc = udf_avg_int4_invalid +); +ERROR: udf_avg_int4_invalid is not its own aggpartialfunc +-- invalid: finalfunc is not equal to serialfunc and stype is internal +-- even though aggpartialfunc is equal to the aggregate name +CREATE AGGREGATE udf_avg_int8_invalid(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + combinefunc = int8_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = udf_avg_int8_invalid +); +ERROR: udf_avg_int8_invalid is not its own aggpartialfunc +-- invalid: aggpartialfunc which doesn't exist +CREATE AGGREGATE nonexistent_aggpartial_agg(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_sum, + combinefunc = int8_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = aggpartial_foo +); +ERROR: function aggpartial_foo(bigint) does not exist +-- invalid: aggpartialfunc is specified and combinefunc isn't specified +CREATE AGGREGATE nocombinefunc_agg(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = avg_p_int8 +); +ERROR: aggcombinefnName must be supplied if aggpartialfnName is supplied +-- invalid: combinefunc in agg and combinefunc in aggpartialfunc don't match +CREATE AGGREGATE udf_avg_invalid_combinefunc(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + combinefunc = numeric_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = avg_p_int8 +); +ERROR: combinefunc in aggregate and combinefunc in its aggpartialfunc must match +CREATE FUNCTION udf_float8_accum(_float8, float8) RETURNS _float8 AS +$$ SELECT float8_accum($1, $2) $$ +LANGUAGE SQL; +-- invalid: sfunc in agg and sfunc in aggpartialfunc don't match +CREATE AGGREGATE udf_avg_invalid_sfunc(float8) ( + sfunc = udf_float8_accum, + stype = _float8, + finalfunc = float8_avg, + combinefunc = float8_combine, + initcond = '{0,0,0}', + aggpartialfunc = avg_p_float8 +); +ERROR: sfunc in aggregate and its aggpartialfunc must match +DROP FUNCTION udf_float8_accum(_float8, float8); +-- invalid: initcond in agg and initcond in aggpartialfunc don't match +CREATE AGGREGATE udf_avg_invalid_initcond(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + initcond = '{1,0}', + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + aggpartialfunc = avg_p_int4 +); +ERROR: initcond in aggregate and initcond in its aggpartialfunc must match +-- invalid: aggserialfn in agg and aggserialfn in aggpartialfunc don't match +CREATE AGGREGATE udf_avg_invalid_aggserialfn(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + combinefunc = int8_avg_combine, + serialfunc = numeric_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = avg_p_int8 +); +ERROR: serialfunc in aggregate and serialfunc in its aggpartialfunc must match +-- invalid: aggdeserialfn in agg and aggdeserialfn in aggpartialfunc don't match +CREATE AGGREGATE udf_avg_invalid_aggdeserialfn(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + combinefunc = int8_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = numeric_avg_deserialize, + aggpartialfunc = avg_p_int8 +); +ERROR: deserialfunc in aggregate and deserialfunc in its aggpartialfunc must match +-- invalid: stype is internal and aggpartialfunc's finalfunc +-- isn't serialfunc of agg +CREATE AGGREGATE udf_avg_p_int8(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_avg_serialize, + combinefunc = int8_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize +); +CREATE AGGREGATE udf_avg(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + combinefunc = int8_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = udf_avg_p_int8 +); +ERROR: finalfunc of aggpartialfunc must match serialfunc of aggregate when stype is internal +-- invalid: stype is not internal and aggpartialfunc has finalfunc +CREATE AGGREGATE udf_avg_p_int4_hasfinal(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + initcond = '{0,0}' +); +CREATE AGGREGATE udf_avg(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + initcond = '{0,0}', + aggpartialfunc = udf_avg_p_int4_hasfinal +); +ERROR: finalfunc of aggpartialfunc must not be supplied when stype isn't internal +-- invalid: current user doesn't have execute privilege on aggpartialfunc +CREATE AGGREGATE udf_avg_p_int4(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + combinefunc = int4_avg_combine, + initcond = '{0,0}' +); +CREATE USER regress_priv_user1; +GRANT ALL ON SCHEMA public TO regress_priv_user1; +REVOKE EXECUTE ON FUNCTION udf_avg_p_int4 FROM public; +SET SESSION AUTHORIZATION regress_priv_user1; +CREATE AGGREGATE udf_avg_noprivilege(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + initcond = '{0,0}', + aggpartialfunc = udf_avg_p_int4 +); +ERROR: permission denied for function udf_avg_p_int4 +SET SESSION AUTHORIZATION DEFAULT; +REVOKE ALL ON SCHEMA public FROM regress_priv_user1; +DROP USER regress_priv_user1; +-- An aggregate function that is aggpartialfunc +-- of another aggregate cannot be deleted. +CREATE AGGREGATE udf_avg_int4_p(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + combinefunc = int4_avg_combine, + initcond = '{0,0}', + aggpartialfunc = udf_avg_int4_p +); +CREATE AGGREGATE udf_avg_int4(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + initcond = '{0,0}', + aggpartialfunc = udf_avg_int4_p +); +DROP AGGREGATE udf_avg_int4_p(int4); +ERROR: cannot drop function udf_avg_int4_p(integer) because other objects depend on it +DETAIL: function udf_avg_int4(integer) depends on function udf_avg_int4_p(integer) +HINT: Use DROP ... CASCADE to drop the dependent objects too. +DROP AGGREGATE udf_avg_int4_p(int4) CASCADE; +NOTICE: drop cascades to function udf_avg_int4(integer) diff --git a/src/test/regress/expected/oidjoins.out b/src/test/regress/expected/oidjoins.out index 215eb899be..6c80efa7eb 100644 --- a/src/test/regress/expected/oidjoins.out +++ b/src/test/regress/expected/oidjoins.out @@ -145,6 +145,7 @@ NOTICE: checking pg_aggregate {aggfinalfn} => pg_proc {oid} NOTICE: checking pg_aggregate {aggcombinefn} => pg_proc {oid} NOTICE: checking pg_aggregate {aggserialfn} => pg_proc {oid} NOTICE: checking pg_aggregate {aggdeserialfn} => pg_proc {oid} +NOTICE: checking pg_aggregate {aggpartialfn} => pg_proc {oid} NOTICE: checking pg_aggregate {aggmtransfn} => pg_proc {oid} NOTICE: checking pg_aggregate {aggminvtransfn} => pg_proc {oid} NOTICE: checking pg_aggregate {aggmfinalfn} => pg_proc {oid} diff --git a/src/test/regress/sql/create_aggregate.sql b/src/test/regress/sql/create_aggregate.sql index d4b4036fd7..e90e0ad2d6 100644 --- a/src/test/regress/sql/create_aggregate.sql +++ b/src/test/regress/sql/create_aggregate.sql @@ -328,3 +328,186 @@ CREATE AGGREGATE case_agg(float8) "Finalfunc_modify" = read_write, "Parallel" = safe ); + +-- invalid: finalfunc is specified and stype is not internal +-- even though aggpartialfunc is equal to the aggregate name +CREATE AGGREGATE udf_avg_int4_invalid(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + initcond = '{0,0}', + aggpartialfunc = udf_avg_int4_invalid +); + +-- invalid: finalfunc is not equal to serialfunc and stype is internal +-- even though aggpartialfunc is equal to the aggregate name +CREATE AGGREGATE udf_avg_int8_invalid(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + combinefunc = int8_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = udf_avg_int8_invalid +); + +-- invalid: aggpartialfunc which doesn't exist +CREATE AGGREGATE nonexistent_aggpartial_agg(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_sum, + combinefunc = int8_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = aggpartial_foo +); + +-- invalid: aggpartialfunc is specified and combinefunc isn't specified +CREATE AGGREGATE nocombinefunc_agg(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = avg_p_int8 +); + +-- invalid: combinefunc in agg and combinefunc in aggpartialfunc don't match +CREATE AGGREGATE udf_avg_invalid_combinefunc(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + combinefunc = numeric_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = avg_p_int8 +); + +CREATE FUNCTION udf_float8_accum(_float8, float8) RETURNS _float8 AS +$$ SELECT float8_accum($1, $2) $$ +LANGUAGE SQL; + +-- invalid: sfunc in agg and sfunc in aggpartialfunc don't match +CREATE AGGREGATE udf_avg_invalid_sfunc(float8) ( + sfunc = udf_float8_accum, + stype = _float8, + finalfunc = float8_avg, + combinefunc = float8_combine, + initcond = '{0,0,0}', + aggpartialfunc = avg_p_float8 +); +DROP FUNCTION udf_float8_accum(_float8, float8); + +-- invalid: initcond in agg and initcond in aggpartialfunc don't match +CREATE AGGREGATE udf_avg_invalid_initcond(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + initcond = '{1,0}', + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + aggpartialfunc = avg_p_int4 +); + +-- invalid: aggserialfn in agg and aggserialfn in aggpartialfunc don't match +CREATE AGGREGATE udf_avg_invalid_aggserialfn(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + combinefunc = int8_avg_combine, + serialfunc = numeric_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = avg_p_int8 +); + +-- invalid: aggdeserialfn in agg and aggdeserialfn in aggpartialfunc don't match +CREATE AGGREGATE udf_avg_invalid_aggdeserialfn(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + combinefunc = int8_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = numeric_avg_deserialize, + aggpartialfunc = avg_p_int8 +); + +-- invalid: stype is internal and aggpartialfunc's finalfunc +-- isn't serialfunc of agg +CREATE AGGREGATE udf_avg_p_int8(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_avg_serialize, + combinefunc = int8_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize +); +CREATE AGGREGATE udf_avg(int8) ( + sfunc = int8_avg_accum, + stype = internal, + finalfunc = numeric_poly_avg, + combinefunc = int8_avg_combine, + serialfunc = int8_avg_serialize, + deserialfunc = int8_avg_deserialize, + aggpartialfunc = udf_avg_p_int8 +); + +-- invalid: stype is not internal and aggpartialfunc has finalfunc +CREATE AGGREGATE udf_avg_p_int4_hasfinal(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + initcond = '{0,0}' +); +CREATE AGGREGATE udf_avg(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + initcond = '{0,0}', + aggpartialfunc = udf_avg_p_int4_hasfinal +); + +-- invalid: current user doesn't have execute privilege on aggpartialfunc +CREATE AGGREGATE udf_avg_p_int4(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + combinefunc = int4_avg_combine, + initcond = '{0,0}' +); +CREATE USER regress_priv_user1; +GRANT ALL ON SCHEMA public TO regress_priv_user1; +REVOKE EXECUTE ON FUNCTION udf_avg_p_int4 FROM public; +SET SESSION AUTHORIZATION regress_priv_user1; + +CREATE AGGREGATE udf_avg_noprivilege(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + initcond = '{0,0}', + aggpartialfunc = udf_avg_p_int4 +); +SET SESSION AUTHORIZATION DEFAULT; +REVOKE ALL ON SCHEMA public FROM regress_priv_user1; +DROP USER regress_priv_user1; + +-- An aggregate function that is aggpartialfunc +-- of another aggregate cannot be deleted. +CREATE AGGREGATE udf_avg_int4_p(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + combinefunc = int4_avg_combine, + initcond = '{0,0}', + aggpartialfunc = udf_avg_int4_p +); +CREATE AGGREGATE udf_avg_int4(int4) ( + sfunc = int4_avg_accum, + stype = _int8, + finalfunc = int8_avg, + combinefunc = int4_avg_combine, + initcond = '{0,0}', + aggpartialfunc = udf_avg_int4_p +); +DROP AGGREGATE udf_avg_int4_p(int4); +DROP AGGREGATE udf_avg_int4_p(int4) CASCADE; -- 2.31.1 ^ permalink raw reply [nested|flat] 56+ messages in thread
* Re: Partial aggregates pushdown 2023-07-10 07:35 RE: Partial aggregates pushdown [email protected] <[email protected]> 2023-07-14 13:40 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]> 2023-07-18 01:35 ` RE: Partial aggregates pushdown [email protected] <[email protected]> 2023-07-19 00:43 ` RE: Partial aggregates pushdown [email protected] <[email protected]> @ 2023-07-20 10:23 ` Alexander Pyhalov <[email protected]> 0 siblings, 0 replies; 56+ messages in thread From: Alexander Pyhalov @ 2023-07-20 10:23 UTC (permalink / raw) To: [email protected] <[email protected]>; +Cc: Bruce Momjian <[email protected]>; pgsql-hackers; Andres Freund <[email protected]>; Tom Lane <[email protected]>; Tomas Vondra <[email protected]>; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; Ilya Gladyshev <[email protected]> [email protected] писал 2023-07-19 03:43: > Hi Mr.Pyhalov, hackers. > 3) > I modified the patch to safely do a partial aggregate pushdown for > queries which contain having clauses. > Hi. Sorry, but I don't see how it could work. For example, the attached test returns wrong result: CREATE FUNCTION f() RETURNS INT AS $$ begin return 10; end $$ LANGUAGE PLPGSQL; SELECT b, sum(a) FROM pagg_tab GROUP BY b HAVING sum(a) < f() ORDER BY 1; b | sum ----+----- 0 | 0 10 | 0 20 | 0 30 | 0 40 | 0 +(5 rows) In fact the above query should have returned 0 rows, as SELECT b, sum(a) FROM pagg_tab GROUP BY b ORDER BY 1; b | sum ----+------ 0 | 600 1 | 660 2 | 720 3 | 780 4 | 840 5 | 900 6 | 960 7 | 1020 8 | 1080 9 | 1140 10 | 600 11 | 660 12 | 720 .... shows no such rows. Or, on the same data SELECT b, sum(a) FROM pagg_tab GROUP BY b HAVING sum(a) > 660 ORDER BY 1; You'll get 0 rows. But SELECT b, sum(a) FROM pagg_tab GROUP BY b; b | sum ----+------ 42 | 720 29 | 1140 4 | 840 34 | 840 41 | 660 0 | 600 40 | 600 gives. The issue is that you can't calculate "partial" having. You should compare full aggregate in filter, but it's not possible on the level of one partition. And you have this in plans Finalize GroupAggregate Output: pagg_tab.b, avg(pagg_tab.a), max(pagg_tab.a), count(*) Group Key: pagg_tab.b Filter: (sum(pagg_tab.a) < 700) -> Sort Output: pagg_tab.b, (PARTIAL avg(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.a)) Sort Key: pagg_tab.b -> Append -> Foreign Scan Output: pagg_tab.b, (PARTIAL avg(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.a)) Filter: ((PARTIAL sum(pagg_tab.a)) < 700) !!!! <--- here we can't compare anything yet, sum is incomplete. Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab) Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*), sum(a) FROM public.pagg_tab_p1 GROUP BY 1 -> Foreign Scan Output: pagg_tab_1.b, (PARTIAL avg(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_1.a)) Filter: ((PARTIAL sum(pagg_tab_1.a)) < 700) Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1) Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*), sum(a) FROM public.pagg_tab_p2 GROUP BY 1 -> Foreign Scan Output: pagg_tab_2.b, (PARTIAL avg(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_2.a)) Filter: ((PARTIAL sum(pagg_tab_2.a)) < 700) Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2) Remote SQL: SELECT b, avg_p_int4(a), max(a), count(*), sum(a) FROM public.pagg_tab_p3 GROUP BY 1 In foreign_grouping_ok() 6586 if (IsA(expr, Aggref)) 6587 { 6588 if (partial) 6589 { 6590 mark_partial_aggref((Aggref *) expr, AGGSPLIT_INITIAL_SERIAL); 6591 continue; 6592 } 6593 else if (!is_foreign_expr(root, grouped_rel, expr)) 6594 return false; 6595 6596 tlist = add_to_flat_tlist(tlist, list_make1(expr)); 6597 } at least you shouldn't do anything with expr, if is_foreign_expr() returned false. If we restrict pushing down queries with havingQuals, I'm not quite sure how Aggref can appear in local_conds. As for changes in planner.c (setGroupClausePartial()) I have several questions. 1) Why don't we add non_group_exprs to pathtarget->exprs when partial_target->exprs is not set? 2) We replace extra->partial_target->exprs with partial_target->exprs after processing. Why are we sure that after this tleSortGroupRef is correct? -- Best regards, Alexander Pyhalov, Postgres Professional Attachments: [text/x-diff] check.patch (1.1K, ../../[email protected]/2-check.patch) download | inline diff: diff --git a/contrib/postgres_fdw/sql/postgres_fdw.sql b/contrib/postgres_fdw/sql/postgres_fdw.sql index 1824cb67fe9..c6f613019c3 100644 --- a/contrib/postgres_fdw/sql/postgres_fdw.sql +++ b/contrib/postgres_fdw/sql/postgres_fdw.sql @@ -3025,6 +3025,22 @@ EXPLAIN (VERBOSE, COSTS OFF) SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1; SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1; +CREATE FUNCTION f() RETURNS INT AS $$ +begin + return 10; +end +$$ LANGUAGE PLPGSQL; + +EXPLAIN (VERBOSE, COSTS OFF) +SELECT b, sum(a) FROM pagg_tab GROUP BY b HAVING sum(a) < f() ORDER BY 1; +SELECT b, sum(a) FROM pagg_tab GROUP BY b HAVING sum(a) < f() ORDER BY 1; + +EXPLAIN (VERBOSE, COSTS OFF) +SELECT b, sum(a) FROM pagg_tab GROUP BY b ORDER BY 1; +SELECT b, sum(a) FROM pagg_tab GROUP BY b ORDER BY 1; + +DROP function f(); + -- Partial aggregates are safe to push down without having clause EXPLAIN (VERBOSE, COSTS OFF) SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1; ^ permalink raw reply [nested|flat] 56+ messages in thread
end of thread, other threads:[~2023-07-20 10:23 UTC | newest] Thread overview: 56+ messages (download: mbox mbox.gz follow: Atom feed) -- links below jump to the message on this page -- 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v2] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v17] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v10] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v8] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v6] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v9] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v7] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v14] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v16] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v11] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v12] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v13] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v18] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v5] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v11] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v15] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2020-02-28 06:52 [PATCH v4] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2022-07-07 02:51 [PATCH v21] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2022-07-07 02:51 [PATCH v19] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2022-07-07 02:51 [PATCH] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2022-07-07 02:51 [PATCH] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2022-07-07 02:51 [PATCH] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2022-11-15 04:41 [PATCH] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2022-11-15 04:41 [PATCH v22] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2022-11-30 02:51 [PATCH v23] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2022-11-30 02:51 [PATCH v24] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2023-03-07 05:55 [PATCH v25] Make End-Of-Recovery error less scary Kyotaro Horiguchi <[email protected]> 2023-07-10 07:35 RE: Partial aggregates pushdown [email protected] <[email protected]> 2023-07-14 13:40 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]> 2023-07-18 01:35 ` RE: Partial aggregates pushdown [email protected] <[email protected]> 2023-07-19 00:43 ` RE: Partial aggregates pushdown [email protected] <[email protected]> 2023-07-20 10:23 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
This inbox is served by agora; see mirroring instructions for how to clone and mirror all data and code used for this inbox