pg.ddx.io pgsql-bugs@postgresql.org mailing list archive help / color / mirror / Atom feed From: Alvaro Herrera <alvherre@alvh.no-ip.org>
To: Kamigishi Rei <iijima.yun@koumakan.jp>
Cc: Andres Freund <andres@anarazel.de>
Cc: Andrew Gierth <andrew@tao11.riddles.org.uk>
Cc: Peter Geoghegan <pg@bowt.ie>
Cc: PostgreSQL mailing lists <pgsql-bugs@lists.postgresql.org>
Cc: Thomas Munro <thomas.munro@gmail.com>
Subject: Re: BUG #17245: Index corruption involving deduplicated entries
Date: Fri, 29 Oct 2021 15:50:32 -0300
Message-ID: <202110291850.cxcsg7j7ye6r@alvherre.pgsql> (raw )
In-Reply-To: <9c7433f6-65b3-9f57-e8a3-4414d0cf0c53@koumakan.jp >
On 2021-Oct-29, Kamigishi Rei wrote:
> The weird part about this is that the WAL archive does not seem to contain
> any data for 157 and 158 above (in 1663/19243/274869 blk 17). The last two
> entries are
>
> rmgr: Btree len (rec/tot): 53/ 4885, tx: 2085600, lsn:
> 2/A0195AE0, prev 2/A01943F0, desc: INSERT_LEAF off 155, blkref #0: rel
> 1663/19243/274869 blk 17 FPW
> rmgr: Btree len (rec/tot): 72/ 72, tx: 2085602, lsn:
> 2/A019DD30, prev 2/A019DCF0, desc: INSERT_LEAF off 156, blkref #0: rel
> 1663/19243/274869 blk 17
>
> The WAL file in data14/pg_wal does not have anything related to 157 and 158
> for this filenode/blk as well.
Hmm, I don't remember precisely how index tuple removal works, but maybe
these 157 and 158 entries existed previously and were truncated away.
It might be useful to have a look at the page header in the page image
contained in the FPW for 2/A0195AE0.
pg_waldump doesn't do it, but I have this patch sitting around ...
probably outdated, but it may be a useful starting point for somebody.
--
Álvaro Herrera Valdivia, Chile — https://www.EnterpriseDB.com/
Attachments:
waldump-save-fpi.patch (text/x-diff, 2.5K) download
diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c
index d1c1cf0fa4..0153af5517 100644
--- a/src/bin/pg_waldump/pg_waldump.c
+++ b/src/bin/pg_waldump/pg_waldump.c
@@ -69,6 +69,8 @@ typedef struct XLogDumpStats
static void fatal_error(const char *fmt,...) pg_attribute_printf(1, 2);
+static XLogRecPtr lsn_to_dump;
+
/*
* Big red button to push when things go horribly wrong.
*/
@@ -445,6 +447,10 @@ XLogDumpDisplayRecord(XLogDumpConfig *config, XLogReaderState *record)
int block_id;
uint8 info = XLogRecGetInfo(record);
XLogRecPtr xl_prev = XLogRecGetPrev(record);
+ bool dump_data = false;
+
+ if (record->ReadRecPtr == lsn_to_dump)
+ dump_data = true;
XLogDumpRecordLen(record, &rec_len, &fpi_len);
@@ -491,6 +497,27 @@ XLogDumpDisplayRecord(XLogDumpConfig *config, XLogReaderState *record)
else
printf(" FPW for WAL verification");
}
+
+ if (dump_data)
+ {
+ char *path;
+ char page[BLCKSZ];
+ FILE *file;
+
+ /* LSN.relfilenode.blkno */
+ path = psprintf("%08x_%08x.%u_%u_%u.%u",
+ (uint32) (record->ReadRecPtr >> 32),
+ (uint32) record->ReadRecPtr,
+ rnode.spcNode, rnode.dbNode, rnode.relNode,
+ blk);
+ file = fopen(path, PG_BINARY_W);
+ if (file == NULL)
+ fatal_error("could not open output file \"%s\": %s\n",
+ path, strerror(errno));
+ RestoreBlockImage(record, block_id, page);
+ fwrite((void *) page, 1, BLCKSZ, file);
+ fclose(file);
+ }
}
putchar('\n');
}
@@ -745,6 +772,7 @@ main(int argc, char **argv)
{"limit", required_argument, NULL, 'n'},
{"path", required_argument, NULL, 'p'},
{"rmgr", required_argument, NULL, 'r'},
+ {"save-fpi", required_argument, NULL, 2},
{"start", required_argument, NULL, 's'},
{"timeline", required_argument, NULL, 't'},
{"xid", required_argument, NULL, 'x'},
@@ -775,6 +803,7 @@ main(int argc, char **argv)
config.filter_by_rmgr = -1;
config.filter_by_xid = InvalidTransactionId;
config.filter_by_xid_enabled = false;
+ lsn_to_dump = InvalidXLogRecPtr;
config.stats = false;
config.stats_per_record = false;
@@ -892,6 +921,17 @@ main(int argc, char **argv)
}
}
break;
+ case 2:
+ /* save-fpi */
+ if (sscanf(optarg, "%X/%X", &xlogid, &xrecoff) != 2)
+ {
+ fprintf(stderr, _("%s: could not parse start WAL location \"%s\"\n"),
+ progname, optarg);
+ goto bad_argument;
+ }
+ else
+ lsn_to_dump = (uint64) xlogid << 32 | xrecoff;
+ break;
default:
goto bad_argument;
}
view thread (103+ messages) latest in thread
Message-ID: <202110291850.cxcsg7j7ye6r@alvherre.pgsql>
Permalink: ../202110291850.cxcsg7j7ye6r@alvherre.pgsql/
Also on: postgresql.org/message-id/202110291850.cxcsg7j7ye6r@alvherre.pgsql
copy link · copy postgr.es · copy git am
reply Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Reply to all the recipients using the --to and --cc options:
reply via email
To: pgsql-bugs@postgresql.org
Cc: alvherre@alvh.no-ip.org, iijima.yun@koumakan.jp, andres@anarazel.de, andrew@tao11.riddles.org.uk, pg@bowt.ie, pgsql-bugs@lists.postgresql.org, thomas.munro@gmail.com
Subject: Re: BUG #17245: Index corruption involving deduplicated entries
In-Reply-To: <202110291850.cxcsg7j7ye6r@alvherre.pgsql>
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox