agora inbox for [email protected]
help / color / mirror / Atom feedRe: [HACKERS] log_destination=file
19+ messages / 10 participants
[nested] [flat]
* Re: [HACKERS] log_destination=file
@ 2017-11-14 15:33 Robert Haas <[email protected]>
2017-11-30 04:44 ` Re: [HACKERS] log_destination=file Michael Paquier <[email protected]>
2018-01-20 12:51 ` Re: [HACKERS] log_destination=file Magnus Hagander <[email protected]>
0 siblings, 2 replies; 19+ messages in thread
From: Robert Haas @ 2017-11-14 15:33 UTC (permalink / raw)
To: Dmitry Dolgov <[email protected]>; +Cc: Magnus Hagander <[email protected]>; Tom Lane <[email protected]>; Greg Stark <[email protected]>; pgsql-hackers
On Sun, Sep 10, 2017 at 5:29 AM, Dmitry Dolgov <[email protected]> wrote:
> average latency:
>
> clients patch master
> 10 0.321 0.286
> 20 0.669 0.602
> 30 1.016 0.942
> 40 1.358 1.280
> 50 1.727 1.637
That's still a noticeable slowdown, though. And we've had previous
reports of the overhead of logging being significant as well:
http://postgr.es/m/CACLsApsA7U0GCFpojVQem6SGTEkv8vnwdBfhVi+dqO+gu5gdCA@mail.gmail.com
I seem to recall a discussion, perhaps in-person, around the time Theo
submitted that patch where it was reported that the logging collector
could not be used on some systems he was working with because it
became a major performance bottleneck. With each backend writing its
own messages to a file, it was tolerable, but when you tried to funnel
everything through a single process, the back-pressure slowed down the
entire system unacceptably.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
^ permalink raw reply [nested|flat] 19+ messages in thread
* Re: [HACKERS] log_destination=file
2017-11-14 15:33 Re: [HACKERS] log_destination=file Robert Haas <[email protected]>
@ 2017-11-30 04:44 ` Michael Paquier <[email protected]>
1 sibling, 0 replies; 19+ messages in thread
From: Michael Paquier @ 2017-11-30 04:44 UTC (permalink / raw)
To: Robert Haas <[email protected]>; +Cc: Dmitry Dolgov <[email protected]>; Magnus Hagander <[email protected]>; Tom Lane <[email protected]>; Greg Stark <[email protected]>; pgsql-hackers
On Wed, Nov 15, 2017 at 12:33 AM, Robert Haas <[email protected]> wrote:
> On Sun, Sep 10, 2017 at 5:29 AM, Dmitry Dolgov <[email protected]> wrote:
>> average latency:
>>
>> clients patch master
>> 10 0.321 0.286
>> 20 0.669 0.602
>> 30 1.016 0.942
>> 40 1.358 1.280
>> 50 1.727 1.637
>
> That's still a noticeable slowdown, though. And we've had previous
> reports of the overhead of logging being significant as well:
>
> http://postgr.es/m/CACLsApsA7U0GCFpojVQem6SGTEkv8vnwdBfhVi+dqO+gu5gdCA@mail.gmail.com
>
> I seem to recall a discussion, perhaps in-person, around the time Theo
> submitted that patch where it was reported that the logging collector
> could not be used on some systems he was working with because it
> became a major performance bottleneck. With each backend writing its
> own messages to a file, it was tolerable, but when you tried to funnel
> everything through a single process, the back-pressure slowed down the
> entire system unacceptably.
I have moved this patch to next CF for now.. The original patch still applies.
--
Michael
^ permalink raw reply [nested|flat] 19+ messages in thread
* Re: [HACKERS] log_destination=file
2017-11-14 15:33 Re: [HACKERS] log_destination=file Robert Haas <[email protected]>
@ 2018-01-20 12:51 ` Magnus Hagander <[email protected]>
2018-01-22 19:52 ` Re: [HACKERS] log_destination=file Robert Haas <[email protected]>
2018-03-02 07:27 ` Re: [HACKERS] log_destination=file Andres Freund <[email protected]>
1 sibling, 2 replies; 19+ messages in thread
From: Magnus Hagander @ 2018-01-20 12:51 UTC (permalink / raw)
To: Robert Haas <[email protected]>; +Cc: Dmitry Dolgov <[email protected]>; Tom Lane <[email protected]>; Greg Stark <[email protected]>; pgsql-hackers
On Tue, Nov 14, 2017 at 5:33 PM, Robert Haas <[email protected]> wrote:
> On Sun, Sep 10, 2017 at 5:29 AM, Dmitry Dolgov <[email protected]>
> wrote:
> > average latency:
> >
> > clients patch master
> > 10 0.321 0.286
> > 20 0.669 0.602
> > 30 1.016 0.942
> > 40 1.358 1.280
> > 50 1.727 1.637
>
> That's still a noticeable slowdown, though. And we've had previous
> reports of the overhead of logging being significant as well:
>
> http://postgr.es/m/CACLsApsA7U0GCFpojVQem6SGTEkv8
> [email protected]
>
> I seem to recall a discussion, perhaps in-person, around the time Theo
> submitted that patch where it was reported that the logging collector
> could not be used on some systems he was working with because it
> became a major performance bottleneck. With each backend writing its
> own messages to a file, it was tolerable, but when you tried to funnel
> everything through a single process, the back-pressure slowed down the
> entire system unacceptably.
>
Finally found myself back at this one, because I still think this is a
problem we definitely need to adress (whether with this file or not).
The funneling into a single process is definitely an issue.
But we don't really solve that problem today wit logging to stderr, do we?
Because somebody has to pick up the log as it came from stderr. Yes, you
get more overhead when sending the log to devnull, but that isn't really a
realistic scenario. The question is what to do when you actually want to
collect that much logging that quickly.
If each backend could actually log to *its own file*, then things would get
sped up. But we can't do that today. Unless you use the hooks and build it
yourself.
Per the thread referenced, using the hooks to handle the
very-high-rate-logging case seems to be the conclusion. But is that still
the conclusion, or do we feel we need to also have a native solution?
And if the conclusion is that hooks is the way to go for that, then is the
slowdown of this patch actually a relevant problem to it?
--
Magnus Hagander
Me: https://www.hagander.net/ <http://www.hagander.net/;
Work: https://www.redpill-linpro.com/ <http://www.redpill-linpro.com/;
^ permalink raw reply [nested|flat] 19+ messages in thread
* Re: [HACKERS] log_destination=file
2017-11-14 15:33 Re: [HACKERS] log_destination=file Robert Haas <[email protected]>
2018-01-20 12:51 ` Re: [HACKERS] log_destination=file Magnus Hagander <[email protected]>
@ 2018-01-22 19:52 ` Robert Haas <[email protected]>
2018-01-22 21:25 ` Re: [HACKERS] log_destination=file Tomas Vondra <[email protected]>
1 sibling, 1 reply; 19+ messages in thread
From: Robert Haas @ 2018-01-22 19:52 UTC (permalink / raw)
To: Magnus Hagander <[email protected]>; +Cc: Dmitry Dolgov <[email protected]>; Tom Lane <[email protected]>; Greg Stark <[email protected]>; pgsql-hackers
On Sat, Jan 20, 2018 at 7:51 AM, Magnus Hagander <[email protected]> wrote:
> Finally found myself back at this one, because I still think this is a
> problem we definitely need to adress (whether with this file or not).
>
> The funneling into a single process is definitely an issue.
>
> But we don't really solve that problem today wit logging to stderr, do we?
> Because somebody has to pick up the log as it came from stderr. Yes, you get
> more overhead when sending the log to devnull, but that isn't really a
> realistic scenario. The question is what to do when you actually want to
> collect that much logging that quickly.
I think it depends on where the bottleneck is. If you're limited by
the speed at which a single process can write, shutting the logging
collector off and letting everyone write fixes it, because now you can
bring the CPU cycles of many processes to bear rather than just one.
If you're limited by the rate at which you can lay the file down on
disk, then turning off the logging collector doesn't help, but I don't
think that's the main problem. Now, of course, if you're writing the
file to disk faster than a single process could do all those writes,
then you're probably also going to need multiple processes to keep up
with reading it, parsing it, etc. But that's not a problem for
PostgreSQL core unless we decide to start shipping an in-core log
analyzer.
> If each backend could actually log to *its own file*, then things would get
> sped up. But we can't do that today. Unless you use the hooks and build it
> yourself.
That seems like a useful thing to support in core.
> Per the thread referenced, using the hooks to handle the
> very-high-rate-logging case seems to be the conclusion. But is that still
> the conclusion, or do we feel we need to also have a native solution?
>
> And if the conclusion is that hooks is the way to go for that, then is the
> slowdown of this patch actually a relevant problem to it?
I think that if we commit what you've proposed, we're making it harder
for people who have a high volume of logging but are not currently
using hooks. I think we should try really hard to avoid the situation
where our suggested workaround for a server change is "go write some C
code and maybe you can get back to the performance you had with
release N-1". That's just not friendly.
I wonder if it would be feasible to set things up so that the logging
collector was always started, but whether or not backends used it or
wrote directly to their original stderr was configurable (e.g. dup
stderr elsewhere, then dup whichever output source is currently
selected onto stderr, then dup the other one if the config is changed
later).
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
^ permalink raw reply [nested|flat] 19+ messages in thread
* Re: [HACKERS] log_destination=file
2017-11-14 15:33 Re: [HACKERS] log_destination=file Robert Haas <[email protected]>
2018-01-20 12:51 ` Re: [HACKERS] log_destination=file Magnus Hagander <[email protected]>
2018-01-22 19:52 ` Re: [HACKERS] log_destination=file Robert Haas <[email protected]>
@ 2018-01-22 21:25 ` Tomas Vondra <[email protected]>
2018-03-02 18:22 ` Re: [HACKERS] log_destination=file Robert Haas <[email protected]>
0 siblings, 1 reply; 19+ messages in thread
From: Tomas Vondra @ 2018-01-22 21:25 UTC (permalink / raw)
To: Robert Haas <[email protected]>; Magnus Hagander <[email protected]>; +Cc: Dmitry Dolgov <[email protected]>; Tom Lane <[email protected]>; Greg Stark <[email protected]>; pgsql-hackers
On 01/22/2018 08:52 PM, Robert Haas wrote:
> On Sat, Jan 20, 2018 at 7:51 AM, Magnus Hagander <[email protected]> wrote:
>> Finally found myself back at this one, because I still think this is a
>> problem we definitely need to adress (whether with this file or not).
>>
>> The funneling into a single process is definitely an issue.
>>
>> But we don't really solve that problem today wit logging to stderr, do we?
>> Because somebody has to pick up the log as it came from stderr. Yes, you get
>> more overhead when sending the log to devnull, but that isn't really a
>> realistic scenario. The question is what to do when you actually want to
>> collect that much logging that quickly.
>
> I think it depends on where the bottleneck is. If you're limited by
> the speed at which a single process can write, shutting the logging
> collector off and letting everyone write fixes it, because now you
> can bring the CPU cycles of many processes to bear rather than just
> one. If you're limited by the rate at which you can lay the file down
> on disk, then turning off the logging collector doesn't help, but I
> don't think that's the main problem. Now, of course, if you're
> writing the file to disk faster than a single process could do all
> those writes, then you're probably also going to need multiple
> processes to keep up with reading it, parsing it, etc. But that's not
> a problem for PostgreSQL core unless we decide to start shipping an
> in-core log analyzer.
>
Sorry for the naive question, but which of these bottlenecks are we
actually hitting? I don't recall dealing with an actual production
system where the log collector would be an issue, so I don't have a very
good idea where the actual bottleneck is in this case.
I find it hard to believe the collector would be limited by I/O when
writing the data to disk (buffered sequential writes and all that).
So I guess it's more about the process of collecting the data from all
the processes through pipe, with the PIPE_MAX_PAYLOAD dance, etc.
I plan to do some experiments with log_statements=all, but my guess is
that the main issue is in transferring individual messages (which may be
further split into multiple chunks). What if we instead send the log
messages in larger batches? Of course, that would require alternative
transport mechanism (say, through shared memory) and the collector would
have to merge to maintain ordering. And then there's the issue that the
collector is pretty critical component.
FWIW I'm pretty sure we're not the only project facing the need to
collect high volume of logs from many processes, so how do the other
projects handle this issue?
>>
>> If each backend could actually log to *its own file*, then things would get
>> sped up. But we can't do that today. Unless you use the hooks and build it
>> yourself.
>
> That seems like a useful thing to support in core.
>
Yeah, I agree with that.
>> Per the thread referenced, using the hooks to handle the
>> very-high-rate-logging case seems to be the conclusion. But is that still
>> the conclusion, or do we feel we need to also have a native solution?
>>
>> And if the conclusion is that hooks is the way to go for that, then is the
>> slowdown of this patch actually a relevant problem to it?
>
> I think that if we commit what you've proposed, we're making it harder
> for people who have a high volume of logging but are not currently
> using hooks. I think we should try really hard to avoid the situation
> where our suggested workaround for a server change is "go write some C
> code and maybe you can get back to the performance you had with
> release N-1". That's just not friendly.
>
> I wonder if it would be feasible to set things up so that the logging
> collector was always started, but whether or not backends used it or
> wrote directly to their original stderr was configurable (e.g. dup
> stderr elsewhere, then dup whichever output source is currently
> selected onto stderr, then dup the other one if the config is changed
> later).
>
I think the hook system is a really powerful tool, but it seems a bit
awkward to force people to use it to improve performance like this ...
That seems like something the core should to out of the box.
regards
--
Tomas Vondra http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
^ permalink raw reply [nested|flat] 19+ messages in thread
* Re: [HACKERS] log_destination=file
2017-11-14 15:33 Re: [HACKERS] log_destination=file Robert Haas <[email protected]>
2018-01-20 12:51 ` Re: [HACKERS] log_destination=file Magnus Hagander <[email protected]>
2018-01-22 19:52 ` Re: [HACKERS] log_destination=file Robert Haas <[email protected]>
2018-01-22 21:25 ` Re: [HACKERS] log_destination=file Tomas Vondra <[email protected]>
@ 2018-03-02 18:22 ` Robert Haas <[email protected]>
2018-03-02 18:29 ` Re: [HACKERS] log_destination=file Tom Lane <[email protected]>
0 siblings, 1 reply; 19+ messages in thread
From: Robert Haas @ 2018-03-02 18:22 UTC (permalink / raw)
To: Tomas Vondra <[email protected]>; +Cc: Magnus Hagander <[email protected]>; Dmitry Dolgov <[email protected]>; Tom Lane <[email protected]>; Greg Stark <[email protected]>; pgsql-hackers
On Mon, Jan 22, 2018 at 4:25 PM, Tomas Vondra
<[email protected]> wrote:
> Sorry for the naive question, but which of these bottlenecks are we
> actually hitting? I don't recall dealing with an actual production
> system where the log collector would be an issue, so I don't have a very
> good idea where the actual bottleneck is in this case.
Unfortunately, neither do I -- but I do believe that the problem is real.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
^ permalink raw reply [nested|flat] 19+ messages in thread
* Re: [HACKERS] log_destination=file
2017-11-14 15:33 Re: [HACKERS] log_destination=file Robert Haas <[email protected]>
2018-01-20 12:51 ` Re: [HACKERS] log_destination=file Magnus Hagander <[email protected]>
2018-01-22 19:52 ` Re: [HACKERS] log_destination=file Robert Haas <[email protected]>
2018-01-22 21:25 ` Re: [HACKERS] log_destination=file Tomas Vondra <[email protected]>
2018-03-02 18:22 ` Re: [HACKERS] log_destination=file Robert Haas <[email protected]>
@ 2018-03-02 18:29 ` Tom Lane <[email protected]>
2018-03-02 23:06 ` Re: [HACKERS] log_destination=file Magnus Hagander <[email protected]>
0 siblings, 1 reply; 19+ messages in thread
From: Tom Lane @ 2018-03-02 18:29 UTC (permalink / raw)
To: Robert Haas <[email protected]>; +Cc: Tomas Vondra <[email protected]>; Magnus Hagander <[email protected]>; Dmitry Dolgov <[email protected]>; Greg Stark <[email protected]>; pgsql-hackers
Robert Haas <[email protected]> writes:
> On Mon, Jan 22, 2018 at 4:25 PM, Tomas Vondra
> <[email protected]> wrote:
>> Sorry for the naive question, but which of these bottlenecks are we
>> actually hitting? I don't recall dealing with an actual production
>> system where the log collector would be an issue, so I don't have a very
>> good idea where the actual bottleneck is in this case.
> Unfortunately, neither do I -- but I do believe that the problem is real.
Not to state the obvious ... but if we do not know where the bottleneck
is, isn't it a bit premature to be proposing changes to "fix" it?
But that argument just leads to the same conclusion that seems to be
prevalent at the moment, ie boot this to v12.
regards, tom lane
^ permalink raw reply [nested|flat] 19+ messages in thread
* Re: [HACKERS] log_destination=file
2017-11-14 15:33 Re: [HACKERS] log_destination=file Robert Haas <[email protected]>
2018-01-20 12:51 ` Re: [HACKERS] log_destination=file Magnus Hagander <[email protected]>
2018-01-22 19:52 ` Re: [HACKERS] log_destination=file Robert Haas <[email protected]>
2018-01-22 21:25 ` Re: [HACKERS] log_destination=file Tomas Vondra <[email protected]>
2018-03-02 18:22 ` Re: [HACKERS] log_destination=file Robert Haas <[email protected]>
2018-03-02 18:29 ` Re: [HACKERS] log_destination=file Tom Lane <[email protected]>
@ 2018-03-02 23:06 ` Magnus Hagander <[email protected]>
0 siblings, 0 replies; 19+ messages in thread
From: Magnus Hagander @ 2018-03-02 23:06 UTC (permalink / raw)
To: Tom Lane <[email protected]>; +Cc: Robert Haas <[email protected]>; Tomas Vondra <[email protected]>; Dmitry Dolgov <[email protected]>; Greg Stark <[email protected]>; pgsql-hackers
On Fri, Mar 2, 2018 at 7:29 PM, Tom Lane <[email protected]> wrote:
> Robert Haas <[email protected]> writes:
> > On Mon, Jan 22, 2018 at 4:25 PM, Tomas Vondra
> > <[email protected]> wrote:
> >> Sorry for the naive question, but which of these bottlenecks are we
> >> actually hitting? I don't recall dealing with an actual production
> >> system where the log collector would be an issue, so I don't have a very
> >> good idea where the actual bottleneck is in this case.
>
> > Unfortunately, neither do I -- but I do believe that the problem is real.
>
> Not to state the obvious ... but if we do not know where the bottleneck
> is, isn't it a bit premature to be proposing changes to "fix" it?
>
> But that argument just leads to the same conclusion that seems to be
> prevalent at the moment, ie boot this to v12.
>
Just to be clear, it has already been booted to v12.
--
Magnus Hagander
Me: https://www.hagander.net/ <http://www.hagander.net/;
Work: https://www.redpill-linpro.com/ <http://www.redpill-linpro.com/;
^ permalink raw reply [nested|flat] 19+ messages in thread
* Re: [HACKERS] log_destination=file
2017-11-14 15:33 Re: [HACKERS] log_destination=file Robert Haas <[email protected]>
2018-01-20 12:51 ` Re: [HACKERS] log_destination=file Magnus Hagander <[email protected]>
@ 2018-03-02 07:27 ` Andres Freund <[email protected]>
2018-03-02 08:42 ` Re: [HACKERS] log_destination=file Magnus Hagander <[email protected]>
1 sibling, 1 reply; 19+ messages in thread
From: Andres Freund @ 2018-03-02 07:27 UTC (permalink / raw)
To: Magnus Hagander <[email protected]>; +Cc: Robert Haas <[email protected]>; Dmitry Dolgov <[email protected]>; Tom Lane <[email protected]>; Greg Stark <[email protected]>; pgsql-hackers
Hi,
On 2018-01-20 14:51:12 +0200, Magnus Hagander wrote:
> Finally found myself back at this one, because I still think this is a
> problem we definitely need to adress (whether with this file or not).
This has an open CF entry for CF 2018-03. Given that there's still a lot
of uncertainty what we want, I don't think that's a reasonable target.
I would suggest moving it to 'returned with feedback'. Arguments
against?
- Andres
^ permalink raw reply [nested|flat] 19+ messages in thread
* Re: [HACKERS] log_destination=file
2017-11-14 15:33 Re: [HACKERS] log_destination=file Robert Haas <[email protected]>
2018-01-20 12:51 ` Re: [HACKERS] log_destination=file Magnus Hagander <[email protected]>
2018-03-02 07:27 ` Re: [HACKERS] log_destination=file Andres Freund <[email protected]>
@ 2018-03-02 08:42 ` Magnus Hagander <[email protected]>
0 siblings, 0 replies; 19+ messages in thread
From: Magnus Hagander @ 2018-03-02 08:42 UTC (permalink / raw)
To: Andres Freund <[email protected]>; +Cc: Robert Haas <[email protected]>; Dmitry Dolgov <[email protected]>; Tom Lane <[email protected]>; Greg Stark <[email protected]>; pgsql-hackers
On Fri, Mar 2, 2018 at 8:27 AM, Andres Freund <[email protected]> wrote:
> Hi,
>
> On 2018-01-20 14:51:12 +0200, Magnus Hagander wrote:
> > Finally found myself back at this one, because I still think this is a
> > problem we definitely need to adress (whether with this file or not).
>
> This has an open CF entry for CF 2018-03. Given that there's still a lot
> of uncertainty what we want, I don't think that's a reasonable target.
> I would suggest moving it to 'returned with feedback'. Arguments
> against?
>
I agree, let's boot it. It needs more discussions about what to actually
do, and after that how. Let's focus on other things for 11, and have this
discussion again when the time comes. I'll go mark it.
--
Magnus Hagander
Me: https://www.hagander.net/ <http://www.hagander.net/;
Work: https://www.redpill-linpro.com/ <http://www.redpill-linpro.com/;
^ permalink raw reply [nested|flat] 19+ messages in thread
* [PATCH v2] Avoid creating archive status ".ready" files too early.
@ 2021-07-30 22:35 Alvaro Herrera <[email protected]>
0 siblings, 0 replies; 19+ messages in thread
From: Alvaro Herrera @ 2021-07-30 22:35 UTC (permalink / raw)
WAL records may span multiple segments, but XLogWrite() does not
wait for the entire record to be written out to disk before
creating archive status files. Instead, as soon as the last WAL
page of the segment is written, the archive status file will be
created. If PostgreSQL crashes before it is able to write the rest
of the record, it will end up reusing segments that have already
been marked as ready-for-archival. However, the archiver process
may have already processed the old version of the segment, so the
wrong version of the segment may be backed-up. This backed-up
segment will cause operations such as point-in-time restores to
fail.
To fix this, we keep track of records that span across segments and
ensure that segments are only marked ready-for-archival once such
records have been completely written to disk.
---
src/backend/access/transam/timeline.c | 2 +-
src/backend/access/transam/xlog.c | 279 ++++++++++++++++++++++-
src/backend/access/transam/xlogarchive.c | 14 +-
src/backend/replication/walreceiver.c | 6 +-
src/backend/storage/lmgr/lwlocknames.txt | 1 +
src/include/access/xlogarchive.h | 4 +-
src/include/access/xlogdefs.h | 3 +
7 files changed, 289 insertions(+), 20 deletions(-)
diff --git a/src/backend/access/transam/timeline.c b/src/backend/access/transam/timeline.c
index 8d0903c175..acd5c2431d 100644
--- a/src/backend/access/transam/timeline.c
+++ b/src/backend/access/transam/timeline.c
@@ -452,7 +452,7 @@ writeTimeLineHistory(TimeLineID newTLI, TimeLineID parentTLI,
if (XLogArchivingActive())
{
TLHistoryFileName(histfname, newTLI);
- XLogArchiveNotify(histfname);
+ XLogArchiveNotify(histfname, true);
}
}
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 26fa2b6c8f..700bbccff8 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -531,6 +531,13 @@ typedef enum ExclusiveBackupState
*/
static SessionBackupState sessionBackupState = SESSION_BACKUP_NONE;
+/* entries for RecordBoundaryMap, used to mark segments ready for archival */
+typedef struct RecordBoundaryEntry
+{
+ XLogSegNo seg; /* must be first */
+ XLogRecPtr pos;
+} RecordBoundaryEntry;
+
/*
* Shared state data for WAL insertion.
*/
@@ -742,6 +749,12 @@ typedef struct XLogCtlData
*/
XLogRecPtr lastFpwDisableRecPtr;
+ /*
+ * The last segment we've marked ready for archival. Protected by
+ * ArchNotifyLock.
+ */
+ XLogSegNo lastNotifiedSeg;
+
slock_t info_lck; /* locks shared variables shown above */
} XLogCtlData;
@@ -755,6 +768,12 @@ static WALInsertLockPadded *WALInsertLocks = NULL;
*/
static ControlFileData *ControlFile = NULL;
+/*
+ * Record boundary map, used for marking segments as ready for archival.
+ * Protected by ArchNotifyLock.
+ */
+static HTAB *RecordBoundaryMap = NULL;
+
/*
* Calculate the amount of space left on the page after 'endptr'. Beware
* multiple evaluation!
@@ -983,6 +1002,12 @@ static XLogRecPtr XLogBytePosToRecPtr(uint64 bytepos);
static XLogRecPtr XLogBytePosToEndRecPtr(uint64 bytepos);
static uint64 XLogRecPtrToBytePos(XLogRecPtr ptr);
static void checkXLogConsistency(XLogReaderState *record);
+static void RegisterRecordBoundaryEntry(XLogSegNo seg, XLogRecPtr pos);
+static void NotifySegmentsReadyForArchive(void);
+static XLogSegNo GetLastNotifiedSegment(void);
+static void SetLastNotifiedSegment(XLogSegNo seg);
+static XLogSegNo GetLatestRecordBoundarySegment(void);
+static void RemoveRecordBoundariesUpTo(XLogSegNo seg);
static void WALInsertLockAcquire(void);
static void WALInsertLockAcquireExclusive(void);
@@ -1030,6 +1055,8 @@ XLogInsertRecord(XLogRecData *rdata,
info == XLOG_SWITCH);
XLogRecPtr StartPos;
XLogRecPtr EndPos;
+ XLogSegNo StartSeg;
+ XLogSegNo EndSeg;
bool prevDoPageWrites = doPageWrites;
/* we assume that all of the record header is in the first chunk */
@@ -1188,6 +1215,31 @@ XLogInsertRecord(XLogRecData *rdata,
SpinLockRelease(&XLogCtl->info_lck);
}
+ /*
+ * Record the record boundary if we crossed the segment boundary. This is
+ * used to ensure that segments are not marked ready for archival before the
+ * entire record has been flushed to disk.
+ *
+ * Note that we do not use XLByteToPrevSeg() for determining the ending
+ * segment. This is done so that a record that fits perfectly into the end
+ * of the segment is marked ready for archival as soon as the flushed
+ * pointer jumps to the next segment.
+ */
+ XLByteToSeg(StartPos, StartSeg, wal_segment_size);
+ XLByteToSeg(EndPos, EndSeg, wal_segment_size);
+
+ if (StartSeg != EndSeg &&
+ XLogArchivingActive())
+ {
+ RegisterRecordBoundaryEntry(EndSeg, EndPos);
+
+ /*
+ * There's a chance that the record was already flushed to disk and we
+ * missed marking segments as ready for archive, so try to do that now.
+ */
+ NotifySegmentsReadyForArchive();
+ }
+
/*
* If this was an XLOG_SWITCH record, flush the record and the empty
* padding space that fills the rest of the segment, and perform
@@ -1285,6 +1337,197 @@ XLogInsertRecord(XLogRecData *rdata,
return EndPos;
}
+/*
+ * RegisterRecordBoundaryEntry
+ *
+ * This enters a new entry into the record boundary map, which is used for
+ * determing when it is safe to mark a segment as ready for archival. An entry
+ * with the given key (the segment number) must not already exist in the map.
+ * Also, the caller is responsible for ensuring that XLByteToSeg() would return
+ * the same segment number for the given record pointer.
+ */
+static void
+RegisterRecordBoundaryEntry(XLogSegNo seg, XLogRecPtr pos)
+{
+ RecordBoundaryEntry *entry;
+ bool found;
+
+ LWLockAcquire(ArchNotifyLock, LW_EXCLUSIVE);
+
+ entry = (RecordBoundaryEntry *) hash_search(RecordBoundaryMap,
+ (void *) &seg, HASH_ENTER,
+ &found);
+ if (found)
+ elog(ERROR, "record boundary entry for segment already exists");
+
+ entry->pos = pos;
+
+ LWLockRelease(ArchNotifyLock);
+}
+
+/*
+ * NotifySegmentsReadyForArchive
+ *
+ * This function marks segments as ready for archival, given that it is safe to
+ * do so. It is safe to call this function repeatedly, even if nothing has
+ * changed since the last time it was called.
+ */
+static void
+NotifySegmentsReadyForArchive(void)
+{
+ XLogRecPtr flushed;
+ XLogSegNo flushed_seg;
+ XLogSegNo latest_boundary_seg;
+ XLogSegNo last_notified;
+
+ /*
+ * We first do a quick sanity check to see if we can bail out without taking
+ * the ArchNotifyLock at all. It is expected that this function will run
+ * frequently and that it will need to do nothing the vast majority of the
+ * time.
+ *
+ * Specifically, we bail out if the shared memory value for the last
+ * notified segment has not yet been initialized or if we've already marked
+ * the segment prior to the segment that contains "flushed" as ready for
+ * archival. We intentionally use XLByteToSeg() instead of
+ * XLByteToPrevSeg() so that we don't skip notifying when a record fits
+ * perfectly into the end of a segment. ("flushed" should point to the
+ * first byte of the record _after_ the one that is known to be flushed to
+ * disk.)
+ */
+ LWLockAcquire(ArchNotifyLock, LW_SHARED);
+ last_notified = GetLastNotifiedSegment();
+ LWLockRelease(ArchNotifyLock);
+
+ if (XLogSegNoIsInvalid(last_notified))
+ return;
+
+ flushed = GetFlushRecPtr();
+ XLByteToSeg(flushed, flushed_seg, wal_segment_size);
+ if (last_notified >= flushed_seg - 1)
+ return;
+
+ /*
+ * At this point, we must acquire ArchNotifyLock before proceeding. In this
+ * section, we look for the latest record boundary in RecordBoundaryMap that
+ * is less than or equal to the current "flushed" pointer, and we notify the
+ * archiver that all segments up to (but not including) that boundary's
+ * associated segment are ready for archival.
+ */
+ LWLockAcquire(ArchNotifyLock, LW_EXCLUSIVE);
+
+ latest_boundary_seg = GetLatestRecordBoundarySegment();
+ if (!XLogSegNoIsInvalid(latest_boundary_seg))
+ {
+ XLogSegNo i;
+
+ /* create the archive status files */
+ for (i = GetLastNotifiedSegment() + 1; i < latest_boundary_seg; i++)
+ XLogArchiveNotifySeg(i, false);
+
+ /* update shared memory */
+ SetLastNotifiedSegment(latest_boundary_seg - 1);
+
+ /* remove old boundaries from the map */
+ RemoveRecordBoundariesUpTo(latest_boundary_seg);
+
+ PgArchWakeup();
+ }
+
+ LWLockRelease(ArchNotifyLock);
+}
+
+/*
+ * GetLatestRecordBoundarySegment
+ *
+ * This function finds the latest record boundary in RecordBoundaryMap that is
+ * less than or equal to the current "flushed" pointer and returns its
+ * associated segment number, given that it is greater than the last notified
+ * segment. Otherwise, InvalidXLogSegNo is returned.
+ *
+ * Caller is expected to be holding ArchNotifyLock.
+ */
+static XLogSegNo
+GetLatestRecordBoundarySegment(void)
+{
+ XLogRecPtr flushed;
+ XLogSegNo flushed_seg;
+ XLogSegNo last_notified;
+
+ flushed = GetFlushRecPtr();
+ XLByteToSeg(flushed, flushed_seg, wal_segment_size);
+ last_notified = GetLastNotifiedSegment();
+
+ for (XLogSegNo i = flushed_seg; i > last_notified; i--)
+ {
+ RecordBoundaryEntry *entry;
+
+ entry = (RecordBoundaryEntry *) hash_search(RecordBoundaryMap,
+ (void *) &i, HASH_FIND,
+ NULL);
+
+ if (entry != NULL && flushed >= entry->pos)
+ return entry->seg;
+ }
+
+ return InvalidXLogSegNo;
+}
+
+/*
+ * RemoveOldRecordBoundaries
+ *
+ * This function removes all entries in the RecordBoundaryMap with segment
+ * numbers up to an including seg.
+ *
+ * Caller is expected to be holding ArchNotifyLock.
+ */
+static void
+RemoveRecordBoundariesUpTo(XLogSegNo seg)
+{
+ RecordBoundaryEntry *entry;
+ HASH_SEQ_STATUS status;
+
+ hash_seq_init(&status, RecordBoundaryMap);
+
+ while ((entry = (RecordBoundaryEntry *) hash_seq_search(&status)) != NULL)
+ {
+ if (entry->seg <= seg)
+ (void) hash_search(RecordBoundaryMap, (void *) &entry->seg,
+ HASH_REMOVE, NULL);
+ }
+}
+
+/*
+ * GetLastNotifiedSegment
+ *
+ * Retrieves last notified segment from shared memory.
+ */
+XLogSegNo
+GetLastNotifiedSegment(void)
+{
+ XLogSegNo seg;
+
+ Assert(LWLockHeldByMe(ArchNotifyLock));
+
+ seg = XLogCtl->lastNotifiedSeg;
+
+ return seg;
+}
+
+/*
+ * SetLastNotifiedSegment
+ *
+ * Sets last notified segment in shared memory. Callers should hold
+ * ArchNotifyLock exclusively when calling this function.
+ */
+static void
+SetLastNotifiedSegment(XLogSegNo seg)
+{
+ Assert(LWLockHeldByMeInMode(ArchNotifyLock, LW_EXCLUSIVE));
+
+ XLogCtl->lastNotifiedSeg = seg;
+}
+
/*
* Reserves the right amount of space for a record of given size from the WAL.
* *StartPos is set to the beginning of the reserved section, *EndPos to
@@ -2607,11 +2850,11 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
* later. Doing it here ensures that one and only one backend will
* perform this fsync.
*
- * This is also the right place to notify the Archiver that the
- * segment is ready to copy to archival storage, and to update the
- * timer for archive_timeout, and to signal for a checkpoint if
- * too many logfile segments have been used since the last
- * checkpoint.
+ * This is also the right place to update the timer for
+ * archive_timeout and to signal for a checkpoint if too many
+ * logfile segments have been used since the last checkpoint. If
+ * lastNotifiedSeg hasn't been initialized yet, we need to do that,
+ * too.
*/
if (finishing_seg)
{
@@ -2623,7 +2866,14 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
LogwrtResult.Flush = LogwrtResult.Write; /* end of page */
if (XLogArchivingActive())
- XLogArchiveNotifySeg(openLogSegNo);
+ {
+ LWLockAcquire(ArchNotifyLock, LW_EXCLUSIVE);
+
+ if (XLogSegNoIsInvalid(GetLastNotifiedSegment()))
+ SetLastNotifiedSegment(openLogSegNo - 1);
+
+ LWLockRelease(ArchNotifyLock);
+ }
XLogCtl->lastSegSwitchTime = (pg_time_t) time(NULL);
XLogCtl->lastSegSwitchLSN = LogwrtResult.Flush;
@@ -2711,6 +2961,9 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
XLogCtl->LogwrtRqst.Flush = LogwrtResult.Flush;
SpinLockRelease(&XLogCtl->info_lck);
}
+
+ if (XLogArchivingActive())
+ NotifySegmentsReadyForArchive();
}
/*
@@ -5140,6 +5393,9 @@ XLOGShmemSize(void)
/* and the buffers themselves */
size = add_size(size, mul_size(XLOG_BLCKSZ, XLOGbuffers));
+ /* stuff for marking segments as ready for archival */
+ size = add_size(size, hash_estimate_size(16, sizeof(RecordBoundaryEntry)));
+
/*
* Note: we don't count ControlFileData, it comes out of the "slop factor"
* added by CreateSharedMemoryAndSemaphores. This lets us use this
@@ -5157,6 +5413,7 @@ XLOGShmemInit(void)
char *allocptr;
int i;
ControlFileData *localControlFile;
+ HASHCTL info;
#ifdef WAL_DEBUG
@@ -5250,12 +5507,20 @@ XLOGShmemInit(void)
XLogCtl->InstallXLogFileSegmentActive = false;
XLogCtl->SharedPromoteIsTriggered = false;
XLogCtl->WalWriterSleeping = false;
+ XLogCtl->lastNotifiedSeg = InvalidXLogSegNo;
SpinLockInit(&XLogCtl->Insert.insertpos_lck);
SpinLockInit(&XLogCtl->info_lck);
SpinLockInit(&XLogCtl->ulsn_lck);
InitSharedLatch(&XLogCtl->recoveryWakeupLatch);
ConditionVariableInit(&XLogCtl->recoveryNotPausedCV);
+
+ /* Initialize stuff for marking segments as ready for archival. */
+ memset(&info, 0, sizeof(info));
+ info.keysize = sizeof(XLogSegNo);
+ info.entrysize = sizeof(RecordBoundaryEntry);
+ RecordBoundaryMap = ShmemInitHash("Record Boundary Table", 16, 16, &info,
+ HASH_ELEM | HASH_BLOBS);
}
/*
@@ -8016,7 +8281,7 @@ StartupXLOG(void)
XLogArchiveCleanup(partialfname);
durable_rename(origpath, partialpath, ERROR);
- XLogArchiveNotify(partialfname);
+ XLogArchiveNotify(partialfname, true);
}
}
}
diff --git a/src/backend/access/transam/xlogarchive.c b/src/backend/access/transam/xlogarchive.c
index 26b023e754..e9cac90b4a 100644
--- a/src/backend/access/transam/xlogarchive.c
+++ b/src/backend/access/transam/xlogarchive.c
@@ -433,7 +433,7 @@ KeepFileRestoredFromArchive(const char *path, const char *xlogfname)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(xlogfname);
else
- XLogArchiveNotify(xlogfname);
+ XLogArchiveNotify(xlogfname, true);
/*
* If the existing file was replaced, since walsenders might have it open,
@@ -464,7 +464,7 @@ KeepFileRestoredFromArchive(const char *path, const char *xlogfname)
* then when complete, rename it to 0000000100000001000000C6.done
*/
void
-XLogArchiveNotify(const char *xlog)
+XLogArchiveNotify(const char *xlog, bool notify)
{
char archiveStatusPath[MAXPGPATH];
FILE *fd;
@@ -489,8 +489,8 @@ XLogArchiveNotify(const char *xlog)
return;
}
- /* Notify archiver that it's got something to do */
- if (IsUnderPostmaster)
+ /* If caller requested, notify archiver that it's got something to do */
+ if (notify)
PgArchWakeup();
}
@@ -498,12 +498,12 @@ XLogArchiveNotify(const char *xlog)
* Convenience routine to notify using segment number representation of filename
*/
void
-XLogArchiveNotifySeg(XLogSegNo segno)
+XLogArchiveNotifySeg(XLogSegNo segno, bool notify)
{
char xlog[MAXFNAMELEN];
XLogFileName(xlog, ThisTimeLineID, segno, wal_segment_size);
- XLogArchiveNotify(xlog);
+ XLogArchiveNotify(xlog, notify);
}
/*
@@ -608,7 +608,7 @@ XLogArchiveCheckDone(const char *xlog)
return true;
/* Retry creation of the .ready file */
- XLogArchiveNotify(xlog);
+ XLogArchiveNotify(xlog, true);
return false;
}
diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c
index 9a2bc37fd7..60de3be92c 100644
--- a/src/backend/replication/walreceiver.c
+++ b/src/backend/replication/walreceiver.c
@@ -622,7 +622,7 @@ WalReceiverMain(void)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(xlogfname);
else
- XLogArchiveNotify(xlogfname);
+ XLogArchiveNotify(xlogfname, true);
}
recvFile = -1;
@@ -760,7 +760,7 @@ WalRcvFetchTimeLineHistoryFiles(TimeLineID first, TimeLineID last)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(fname);
else
- XLogArchiveNotify(fname);
+ XLogArchiveNotify(fname, true);
pfree(fname);
pfree(content);
@@ -915,7 +915,7 @@ XLogWalRcvWrite(char *buf, Size nbytes, XLogRecPtr recptr)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(xlogfname);
else
- XLogArchiveNotify(xlogfname);
+ XLogArchiveNotify(xlogfname, true);
}
recvFile = -1;
diff --git a/src/backend/storage/lmgr/lwlocknames.txt b/src/backend/storage/lmgr/lwlocknames.txt
index 6c7cf6c295..d39225bf94 100644
--- a/src/backend/storage/lmgr/lwlocknames.txt
+++ b/src/backend/storage/lmgr/lwlocknames.txt
@@ -53,3 +53,4 @@ XactTruncationLock 44
# 45 was XactTruncationLock until removal of BackendRandomLock
WrapLimitsVacuumLock 46
NotifyQueueTailLock 47
+ArchNotifyLock 48
diff --git a/src/include/access/xlogarchive.h b/src/include/access/xlogarchive.h
index 3edd1a976c..e25c3d8117 100644
--- a/src/include/access/xlogarchive.h
+++ b/src/include/access/xlogarchive.h
@@ -23,8 +23,8 @@ extern bool RestoreArchivedFile(char *path, const char *xlogfname,
extern void ExecuteRecoveryCommand(const char *command, const char *commandName,
bool failOnSignal);
extern void KeepFileRestoredFromArchive(const char *path, const char *xlogfname);
-extern void XLogArchiveNotify(const char *xlog);
-extern void XLogArchiveNotifySeg(XLogSegNo segno);
+extern void XLogArchiveNotify(const char *xlog, bool notify);
+extern void XLogArchiveNotifySeg(XLogSegNo segno, bool notify);
extern void XLogArchiveForceDone(const char *xlog);
extern bool XLogArchiveCheckDone(const char *xlog);
extern bool XLogArchiveIsBusy(const char *xlog);
diff --git a/src/include/access/xlogdefs.h b/src/include/access/xlogdefs.h
index 60348d1850..8d172f887a 100644
--- a/src/include/access/xlogdefs.h
+++ b/src/include/access/xlogdefs.h
@@ -47,6 +47,9 @@ typedef uint64 XLogRecPtr;
*/
typedef uint64 XLogSegNo;
+#define InvalidXLogSegNo 0xFFFFFFFFFFFFFFFF
+#define XLogSegNoIsInvalid(s) ((s) == InvalidXLogSegNo)
+
/*
* TimeLineID (TLI) - identifies different database histories to prevent
* confusion after restoring a prior state of a database installation.
--
2.20.1
--mdy6umzbb7bbupyj--
^ permalink raw reply [nested|flat] 19+ messages in thread
* [PATCH v3] Avoid creating archive status ".ready" files too early.
@ 2021-07-30 22:35 Alvaro Herrera <[email protected]>
0 siblings, 0 replies; 19+ messages in thread
From: Alvaro Herrera @ 2021-07-30 22:35 UTC (permalink / raw)
WAL records may span multiple segments, but XLogWrite() does not
wait for the entire record to be written out to disk before
creating archive status files. Instead, as soon as the last WAL
page of the segment is written, the archive status file will be
created. If PostgreSQL crashes before it is able to write the rest
of the record, it will end up reusing segments that have already
been marked as ready-for-archival. However, the archiver process
may have already processed the old version of the segment, so the
wrong version of the segment may be backed-up. This backed-up
segment will cause operations such as point-in-time restores to
fail.
To fix this, we keep track of records that span across segments and
ensure that segments are only marked ready-for-archival once such
records have been completely written to disk.
---
src/backend/access/transam/timeline.c | 2 +-
src/backend/access/transam/xlog.c | 286 ++++++++++++++++++++++-
src/backend/access/transam/xlogarchive.c | 14 +-
src/backend/replication/walreceiver.c | 6 +-
src/backend/storage/lmgr/lwlocknames.txt | 1 +
src/include/access/xlogarchive.h | 4 +-
src/include/access/xlogdefs.h | 5 +
7 files changed, 296 insertions(+), 22 deletions(-)
diff --git a/src/backend/access/transam/timeline.c b/src/backend/access/transam/timeline.c
index 8d0903c175..acd5c2431d 100644
--- a/src/backend/access/transam/timeline.c
+++ b/src/backend/access/transam/timeline.c
@@ -452,7 +452,7 @@ writeTimeLineHistory(TimeLineID newTLI, TimeLineID parentTLI,
if (XLogArchivingActive())
{
TLHistoryFileName(histfname, newTLI);
- XLogArchiveNotify(histfname);
+ XLogArchiveNotify(histfname, true);
}
}
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f84c0bb01e..99fe1ac0a2 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -512,6 +512,13 @@ typedef enum ExclusiveBackupState
*/
static SessionBackupState sessionBackupState = SESSION_BACKUP_NONE;
+/* entries for RecordBoundaryMap, used to mark segments ready for archival */
+typedef struct RecordBoundaryEntry
+{
+ XLogSegNo seg; /* must be first */
+ XLogRecPtr pos;
+} RecordBoundaryEntry;
+
/*
* Shared state data for WAL insertion.
*/
@@ -723,6 +730,12 @@ typedef struct XLogCtlData
*/
XLogRecPtr lastFpwDisableRecPtr;
+ /*
+ * The last segment we've marked ready for archival. Protected by
+ * info_lck.
+ */
+ XLogSegNo lastNotifiedSeg;
+
slock_t info_lck; /* locks shared variables shown above */
} XLogCtlData;
@@ -736,6 +749,12 @@ static WALInsertLockPadded *WALInsertLocks = NULL;
*/
static ControlFileData *ControlFile = NULL;
+/*
+ * Record boundary map, used for marking segments as ready for archival.
+ * Protected by ArchNotifyLock.
+ */
+static HTAB *RecordBoundaryMap = NULL;
+
/*
* Calculate the amount of space left on the page after 'endptr'. Beware
* multiple evaluation!
@@ -962,6 +981,13 @@ static XLogRecPtr XLogBytePosToRecPtr(uint64 bytepos);
static XLogRecPtr XLogBytePosToEndRecPtr(uint64 bytepos);
static uint64 XLogRecPtrToBytePos(XLogRecPtr ptr);
static void checkXLogConsistency(XLogReaderState *record);
+static void RegisterRecordBoundaryEntry(XLogSegNo seg, XLogRecPtr pos);
+static void NotifySegmentsReadyForArchive(void);
+static XLogSegNo GetLastNotifiedSegment(void);
+static void SetLastNotifiedSegment(XLogSegNo seg);
+static void SetLastNotifiedSegmentIfInvalid(XLogSegNo seg);
+static XLogSegNo GetLatestRecordBoundarySegment(void);
+static void RemoveRecordBoundariesUpTo(XLogSegNo seg);
static void WALInsertLockAcquire(void);
static void WALInsertLockAcquireExclusive(void);
@@ -1009,6 +1035,8 @@ XLogInsertRecord(XLogRecData *rdata,
info == XLOG_SWITCH);
XLogRecPtr StartPos;
XLogRecPtr EndPos;
+ XLogSegNo StartSeg;
+ XLogSegNo EndSeg;
bool prevDoPageWrites = doPageWrites;
/* we assume that all of the record header is in the first chunk */
@@ -1167,10 +1195,33 @@ XLogInsertRecord(XLogRecData *rdata,
SpinLockRelease(&XLogCtl->info_lck);
}
+ /*
+ * Record the record boundary if we crossed the segment boundary. This is
+ * used to ensure that segments are not marked ready for archival before the
+ * entire record has been flushed to disk.
+ *
+ * Note that we do not use XLByteToPrevSeg() for determining the ending
+ * segment. This is done so that a record that fits perfectly into the end
+ * of the segment is marked ready for archival as soon as the flushed
+ * pointer jumps to the next segment.
+ */
+ XLByteToSeg(StartPos, StartSeg, wal_segment_size);
+ XLByteToSeg(EndPos, EndSeg, wal_segment_size);
+
+ if (StartSeg != EndSeg && XLogArchivingActive())
+ {
+ RegisterRecordBoundaryEntry(EndSeg, EndPos);
+
+ /*
+ * There's a chance that the record was already flushed to disk and we
+ * missed marking segments as ready for archive, so try to do that now.
+ */
+ NotifySegmentsReadyForArchive();
+ }
+
/*
* If this was an XLOG_SWITCH record, flush the record and the empty
- * padding space that fills the rest of the segment, and perform
- * end-of-segment actions (eg, notifying archiver).
+ * padding space that fills the rest of the segment.
*/
if (isLogSwitch)
{
@@ -1264,6 +1315,205 @@ XLogInsertRecord(XLogRecData *rdata,
return EndPos;
}
+/*
+ * RegisterRecordBoundaryEntry
+ *
+ * This enters a new entry into the record boundary map, which is used for
+ * determing when it is safe to mark a segment as ready for archival. An entry
+ * with the given key (the segment number) must not already exist in the map.
+ * Also, the caller is responsible for ensuring that XLByteToSeg() would return
+ * the same segment number for the given record pointer.
+ */
+static void
+RegisterRecordBoundaryEntry(XLogSegNo seg, XLogRecPtr pos)
+{
+ RecordBoundaryEntry *entry;
+ bool found;
+
+ LWLockAcquire(ArchNotifyLock, LW_EXCLUSIVE);
+
+ entry = (RecordBoundaryEntry *) hash_search(RecordBoundaryMap,
+ (void *) &seg, HASH_ENTER,
+ &found);
+ if (found)
+ elog(ERROR, "record boundary entry for segment already exists");
+
+ entry->pos = pos;
+
+ LWLockRelease(ArchNotifyLock);
+}
+
+/*
+ * NotifySegmentsReadyForArchive
+ *
+ * This function marks segments as ready for archival, given that it is safe to
+ * do so. It is safe to call this function repeatedly, even if nothing has
+ * changed since the last time it was called.
+ */
+static void
+NotifySegmentsReadyForArchive(void)
+{
+ XLogRecPtr flushed;
+ XLogSegNo flushed_seg;
+ XLogSegNo latest_boundary_seg;
+ XLogSegNo last_notified;
+
+ /*
+ * We first do a quick sanity check to see if we can bail out without taking
+ * the ArchNotifyLock at all. It is expected that this function will run
+ * frequently and that it will need to do nothing the vast majority of the
+ * time.
+ *
+ * Specifically, we bail out if the shared memory value for the last
+ * notified segment has not yet been initialized or if we've already marked
+ * the segment prior to the segment that contains "flushed" as ready for
+ * archival. We intentionally use XLByteToSeg() instead of
+ * XLByteToPrevSeg() so that we don't skip notifying when a record fits
+ * perfectly into the end of a segment. ("flushed" should point to the
+ * first byte of the record _after_ the one that is known to be flushed to
+ * disk.)
+ */
+ last_notified = GetLastNotifiedSegment();
+
+ if (XLogSegNoIsInvalid(last_notified))
+ return;
+
+ flushed = GetFlushRecPtr();
+ XLByteToSeg(flushed, flushed_seg, wal_segment_size);
+ if (last_notified >= flushed_seg - 1)
+ return;
+
+ /*
+ * Notify archiver about segments that are ready for archival (by creating
+ * the corresponding .ready files), and discard segment boundaries no
+ * longer needed.
+ */
+ LWLockAcquire(ArchNotifyLock, LW_EXCLUSIVE);
+ latest_boundary_seg = GetLatestRecordBoundarySegment();
+ if (!XLogSegNoIsInvalid(latest_boundary_seg))
+ {
+ XLogSegNo seg;
+
+ /* create the archive status files */
+ for (seg = GetLastNotifiedSegment() + 1; seg < latest_boundary_seg; seg++)
+ XLogArchiveNotifySeg(seg, false);
+
+ /* update shared memory */
+ SetLastNotifiedSegment(latest_boundary_seg - 1);
+
+ /* remove old boundaries from the map */
+ RemoveRecordBoundariesUpTo(latest_boundary_seg);
+
+ PgArchWakeup();
+ }
+ LWLockRelease(ArchNotifyLock);
+}
+
+/*
+ * GetLatestRecordBoundarySegment
+ *
+ * This function finds the latest record boundary in RecordBoundaryMap that is
+ * less than or equal to the current "flushed" pointer and returns its
+ * associated segment number, given that it is greater than the last notified
+ * segment. Otherwise, InvalidXLogSegNo is returned.
+ */
+static XLogSegNo
+GetLatestRecordBoundarySegment(void)
+{
+ XLogRecPtr flushed;
+ XLogSegNo flushed_seg;
+ XLogSegNo last_notified;
+ XLogSegNo seg;
+
+ Assert(LWLockHeldByMe(ArchNotifyLock));
+
+ flushed = GetFlushRecPtr();
+ XLByteToSeg(flushed, flushed_seg, wal_segment_size);
+ last_notified = GetLastNotifiedSegment();
+
+ for (seg = flushed_seg; seg > last_notified; seg--)
+ {
+ RecordBoundaryEntry *entry;
+
+ entry = (RecordBoundaryEntry *) hash_search(RecordBoundaryMap,
+ (void *) &seg, HASH_FIND,
+ NULL);
+
+ if (entry != NULL && flushed >= entry->pos)
+ return entry->seg;
+ }
+
+ return InvalidXLogSegNo;
+}
+
+/*
+ * RemoveOldRecordBoundaries
+ *
+ * This function removes all entries in the RecordBoundaryMap with segment
+ * numbers up to an including seg.
+ */
+static void
+RemoveRecordBoundariesUpTo(XLogSegNo seg)
+{
+ RecordBoundaryEntry *entry;
+ HASH_SEQ_STATUS status;
+
+ Assert(LWLockHeldByMeInMode(ArchNotifyLock, LW_EXCLUSIVE));
+
+ hash_seq_init(&status, RecordBoundaryMap);
+
+ while ((entry = (RecordBoundaryEntry *) hash_seq_search(&status)) != NULL)
+ {
+ if (entry->seg <= seg)
+ (void) hash_search(RecordBoundaryMap, (void *) &entry->seg,
+ HASH_REMOVE, NULL);
+ }
+}
+
+/*
+ * GetLastNotifiedSegment
+ *
+ * Retrieves last notified segment from shared memory.
+ */
+XLogSegNo
+GetLastNotifiedSegment(void)
+{
+ XLogSegNo seg;
+
+ SpinLockAcquire(&XLogCtl->info_lck);
+ seg = XLogCtl->lastNotifiedSeg;
+ SpinLockRelease(&XLogCtl->info_lck);
+
+ return seg;
+}
+
+/*
+ * SetLastNotifiedSegment
+ *
+ * Sets last notified segment in shared memory.
+ */
+static void
+SetLastNotifiedSegment(XLogSegNo seg)
+{
+ SpinLockAcquire(&XLogCtl->info_lck);
+ XLogCtl->lastNotifiedSeg = seg;
+ SpinLockRelease(&XLogCtl->info_lck);
+}
+
+/*
+ * SetLastNotifiedSegmentIfInvalid
+ *
+ * Sets last notified segment, but only if it's currently unset.
+ */
+static void
+SetLastNotifiedSegmentIfInvalid(XLogSegNo seg)
+{
+ SpinLockAcquire(&XLogCtl->info_lck);
+ if (XLogCtl->lastNotifiedSeg == InvalidXLogSegNo)
+ XLogCtl->lastNotifiedSeg = seg;
+ SpinLockRelease(&XLogCtl->info_lck);
+}
+
/*
* Reserves the right amount of space for a record of given size from the WAL.
* *StartPos is set to the beginning of the reserved section, *EndPos to
@@ -2586,11 +2836,14 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
* later. Doing it here ensures that one and only one backend will
* perform this fsync.
*
- * This is also the right place to notify the Archiver that the
- * segment is ready to copy to archival storage, and to update the
- * timer for archive_timeout, and to signal for a checkpoint if
- * too many logfile segments have been used since the last
- * checkpoint.
+ * If WAL archiving is active and lastNotifiedSeg hasn't been
+ * initialized yet, do that, too. This will let us notify
+ * archiver correctly later.
+ *
+ * This is also the right place to update the timer for
+ * archive_timeout and to signal for a checkpoint if too many
+ * logfile segments have been used since the last checkpoint.
+ *
*/
if (finishing_seg)
{
@@ -2602,7 +2855,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
LogwrtResult.Flush = LogwrtResult.Write; /* end of page */
if (XLogArchivingActive())
- XLogArchiveNotifySeg(openLogSegNo);
+ SetLastNotifiedSegmentIfInvalid(openLogSegNo - 1);
XLogCtl->lastSegSwitchTime = (pg_time_t) time(NULL);
XLogCtl->lastSegSwitchLSN = LogwrtResult.Flush;
@@ -2690,6 +2943,9 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
XLogCtl->LogwrtRqst.Flush = LogwrtResult.Flush;
SpinLockRelease(&XLogCtl->info_lck);
}
+
+ if (XLogArchivingActive())
+ NotifySegmentsReadyForArchive();
}
/*
@@ -5117,6 +5373,9 @@ XLOGShmemSize(void)
/* and the buffers themselves */
size = add_size(size, mul_size(XLOG_BLCKSZ, XLOGbuffers));
+ /* stuff for marking segments as ready for archival */
+ size = add_size(size, hash_estimate_size(16, sizeof(RecordBoundaryEntry)));
+
/*
* Note: we don't count ControlFileData, it comes out of the "slop factor"
* added by CreateSharedMemoryAndSemaphores. This lets us use this
@@ -5134,6 +5393,7 @@ XLOGShmemInit(void)
char *allocptr;
int i;
ControlFileData *localControlFile;
+ HASHCTL info;
#ifdef WAL_DEBUG
@@ -5227,12 +5487,20 @@ XLOGShmemInit(void)
XLogCtl->InstallXLogFileSegmentActive = false;
XLogCtl->SharedPromoteIsTriggered = false;
XLogCtl->WalWriterSleeping = false;
+ XLogCtl->lastNotifiedSeg = InvalidXLogSegNo;
SpinLockInit(&XLogCtl->Insert.insertpos_lck);
SpinLockInit(&XLogCtl->info_lck);
SpinLockInit(&XLogCtl->ulsn_lck);
InitSharedLatch(&XLogCtl->recoveryWakeupLatch);
ConditionVariableInit(&XLogCtl->recoveryNotPausedCV);
+
+ /* Initialize stuff for marking segments as ready for archival. */
+ memset(&info, 0, sizeof(info));
+ info.keysize = sizeof(XLogSegNo);
+ info.entrysize = sizeof(RecordBoundaryEntry);
+ RecordBoundaryMap = ShmemInitHash("Record Boundary Table", 16, 16, &info,
+ HASH_ELEM | HASH_BLOBS);
}
/*
@@ -7997,7 +8265,7 @@ StartupXLOG(void)
XLogArchiveCleanup(partialfname);
durable_rename(origpath, partialpath, ERROR);
- XLogArchiveNotify(partialfname);
+ XLogArchiveNotify(partialfname, true);
}
}
}
diff --git a/src/backend/access/transam/xlogarchive.c b/src/backend/access/transam/xlogarchive.c
index 26b023e754..e9cac90b4a 100644
--- a/src/backend/access/transam/xlogarchive.c
+++ b/src/backend/access/transam/xlogarchive.c
@@ -433,7 +433,7 @@ KeepFileRestoredFromArchive(const char *path, const char *xlogfname)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(xlogfname);
else
- XLogArchiveNotify(xlogfname);
+ XLogArchiveNotify(xlogfname, true);
/*
* If the existing file was replaced, since walsenders might have it open,
@@ -464,7 +464,7 @@ KeepFileRestoredFromArchive(const char *path, const char *xlogfname)
* then when complete, rename it to 0000000100000001000000C6.done
*/
void
-XLogArchiveNotify(const char *xlog)
+XLogArchiveNotify(const char *xlog, bool notify)
{
char archiveStatusPath[MAXPGPATH];
FILE *fd;
@@ -489,8 +489,8 @@ XLogArchiveNotify(const char *xlog)
return;
}
- /* Notify archiver that it's got something to do */
- if (IsUnderPostmaster)
+ /* If caller requested, notify archiver that it's got something to do */
+ if (notify)
PgArchWakeup();
}
@@ -498,12 +498,12 @@ XLogArchiveNotify(const char *xlog)
* Convenience routine to notify using segment number representation of filename
*/
void
-XLogArchiveNotifySeg(XLogSegNo segno)
+XLogArchiveNotifySeg(XLogSegNo segno, bool notify)
{
char xlog[MAXFNAMELEN];
XLogFileName(xlog, ThisTimeLineID, segno, wal_segment_size);
- XLogArchiveNotify(xlog);
+ XLogArchiveNotify(xlog, notify);
}
/*
@@ -608,7 +608,7 @@ XLogArchiveCheckDone(const char *xlog)
return true;
/* Retry creation of the .ready file */
- XLogArchiveNotify(xlog);
+ XLogArchiveNotify(xlog, true);
return false;
}
diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c
index 9a2bc37fd7..60de3be92c 100644
--- a/src/backend/replication/walreceiver.c
+++ b/src/backend/replication/walreceiver.c
@@ -622,7 +622,7 @@ WalReceiverMain(void)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(xlogfname);
else
- XLogArchiveNotify(xlogfname);
+ XLogArchiveNotify(xlogfname, true);
}
recvFile = -1;
@@ -760,7 +760,7 @@ WalRcvFetchTimeLineHistoryFiles(TimeLineID first, TimeLineID last)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(fname);
else
- XLogArchiveNotify(fname);
+ XLogArchiveNotify(fname, true);
pfree(fname);
pfree(content);
@@ -915,7 +915,7 @@ XLogWalRcvWrite(char *buf, Size nbytes, XLogRecPtr recptr)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(xlogfname);
else
- XLogArchiveNotify(xlogfname);
+ XLogArchiveNotify(xlogfname, true);
}
recvFile = -1;
diff --git a/src/backend/storage/lmgr/lwlocknames.txt b/src/backend/storage/lmgr/lwlocknames.txt
index 6c7cf6c295..d39225bf94 100644
--- a/src/backend/storage/lmgr/lwlocknames.txt
+++ b/src/backend/storage/lmgr/lwlocknames.txt
@@ -53,3 +53,4 @@ XactTruncationLock 44
# 45 was XactTruncationLock until removal of BackendRandomLock
WrapLimitsVacuumLock 46
NotifyQueueTailLock 47
+ArchNotifyLock 48
diff --git a/src/include/access/xlogarchive.h b/src/include/access/xlogarchive.h
index 3edd1a976c..e25c3d8117 100644
--- a/src/include/access/xlogarchive.h
+++ b/src/include/access/xlogarchive.h
@@ -23,8 +23,8 @@ extern bool RestoreArchivedFile(char *path, const char *xlogfname,
extern void ExecuteRecoveryCommand(const char *command, const char *commandName,
bool failOnSignal);
extern void KeepFileRestoredFromArchive(const char *path, const char *xlogfname);
-extern void XLogArchiveNotify(const char *xlog);
-extern void XLogArchiveNotifySeg(XLogSegNo segno);
+extern void XLogArchiveNotify(const char *xlog, bool notify);
+extern void XLogArchiveNotifySeg(XLogSegNo segno, bool notify);
extern void XLogArchiveForceDone(const char *xlog);
extern bool XLogArchiveCheckDone(const char *xlog);
extern bool XLogArchiveIsBusy(const char *xlog);
diff --git a/src/include/access/xlogdefs.h b/src/include/access/xlogdefs.h
index 60348d1850..b5a0023d81 100644
--- a/src/include/access/xlogdefs.h
+++ b/src/include/access/xlogdefs.h
@@ -47,6 +47,11 @@ typedef uint64 XLogRecPtr;
*/
typedef uint64 XLogSegNo;
+#define InvalidXLogSegNo ((XLogSegNo) 0xFFFFFFFFFFFFFFFF)
+#define XLogSegNoIsInvalid(s) \
+ (AssertVariableIsOfTypeMacro(s, XLogSegNo), \
+ (s) == InvalidXLogSegNo)
+
/*
* TimeLineID (TLI) - identifies different database histories to prevent
* confusion after restoring a prior state of a database installation.
--
2.20.1
--vp4c52all22p7wxm--
^ permalink raw reply [nested|flat] 19+ messages in thread
* [PATCH v10] Avoid creating archive status ".ready" files too early.
@ 2021-08-17 03:52 Nathan Bossart <[email protected]>
0 siblings, 0 replies; 19+ messages in thread
From: Nathan Bossart @ 2021-08-17 03:52 UTC (permalink / raw)
WAL records may span multiple segments, but XLogWrite() does not
wait for the entire record to be written out to disk before
creating archive status files. Instead, as soon as the last WAL
page of the segment is written, the archive status file will be
created. If PostgreSQL crashes before it is able to write the rest
of the record, it will end up reusing segments that have already
been marked as ready-for-archival. However, the archiver process
may have already processed the old version of the segment, so the
wrong version of the segment may be backed-up. This backed-up
segment will cause operations such as point-in-time restores to
fail.
To fix this, we keep track of records that span across segments and
ensure that segments are only marked ready-for-archival once such
records have been completely written to disk.
---
src/backend/access/transam/timeline.c | 2 +-
src/backend/access/transam/xlog.c | 291 ++++++++++++++++++++++-
src/backend/access/transam/xlogarchive.c | 17 +-
src/backend/postmaster/walwriter.c | 7 +
src/backend/replication/walreceiver.c | 6 +-
src/backend/storage/lmgr/lwlocknames.txt | 1 +
src/include/access/xlog.h | 1 +
src/include/access/xlogarchive.h | 4 +-
src/include/access/xlogdefs.h | 1 +
src/tools/pgindent/typedefs.list | 1 +
10 files changed, 309 insertions(+), 22 deletions(-)
diff --git a/src/backend/access/transam/timeline.c b/src/backend/access/transam/timeline.c
index 8d0903c175..acd5c2431d 100644
--- a/src/backend/access/transam/timeline.c
+++ b/src/backend/access/transam/timeline.c
@@ -452,7 +452,7 @@ writeTimeLineHistory(TimeLineID newTLI, TimeLineID parentTLI,
if (XLogArchivingActive())
{
TLHistoryFileName(histfname, newTLI);
- XLogArchiveNotify(histfname);
+ XLogArchiveNotify(histfname, true);
}
}
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e51a7a749d..d2ccf2a7bb 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -512,6 +512,16 @@ typedef enum ExclusiveBackupState
*/
static SessionBackupState sessionBackupState = SESSION_BACKUP_NONE;
+/*
+ * Entries for SegmentBoundaryMap. Each such entry represents a WAL
+ * record that ends in endpos and crosses a WAL segment boundary.
+ */
+typedef struct SegmentBoundaryEntry
+{
+ XLogSegNo seg; /* must be first */
+ XLogRecPtr endpos;
+} SegmentBoundaryEntry;
+
/*
* Shared state data for WAL insertion.
*/
@@ -723,6 +733,12 @@ typedef struct XLogCtlData
*/
XLogRecPtr lastFpwDisableRecPtr;
+ /*
+ * The last segment we've marked ready for archival. Protected by
+ * info_lck.
+ */
+ XLogSegNo lastNotifiedSeg;
+
slock_t info_lck; /* locks shared variables shown above */
} XLogCtlData;
@@ -736,6 +752,12 @@ static WALInsertLockPadded *WALInsertLocks = NULL;
*/
static ControlFileData *ControlFile = NULL;
+/*
+ * Segment boundary map, used for marking segments as ready for archival.
+ * Protected by SegmentBoundaryLock.
+ */
+static HTAB *SegmentBoundaryMap = NULL;
+
/*
* Calculate the amount of space left on the page after 'endptr'. Beware
* multiple evaluation!
@@ -962,6 +984,13 @@ static XLogRecPtr XLogBytePosToRecPtr(uint64 bytepos);
static XLogRecPtr XLogBytePosToEndRecPtr(uint64 bytepos);
static uint64 XLogRecPtrToBytePos(XLogRecPtr ptr);
static void checkXLogConsistency(XLogReaderState *record);
+static void RegisterSegmentBoundaryEntry(XLogSegNo seg, XLogRecPtr pos);
+static XLogSegNo GetLastNotifiedSegment(void);
+static void SetLastNotifiedSegment(XLogSegNo seg);
+static bool GetLatestSegmentBoundary(XLogSegNo last_notified,
+ XLogRecPtr flushed,
+ XLogSegNo *latest_boundary_seg);
+static void RemoveSegmentBoundariesUpTo(XLogSegNo seg);
static void WALInsertLockAcquire(void);
static void WALInsertLockAcquireExclusive(void);
@@ -1158,6 +1187,9 @@ XLogInsertRecord(XLogRecData *rdata,
*/
if (StartPos / XLOG_BLCKSZ != EndPos / XLOG_BLCKSZ)
{
+ XLogSegNo StartSeg;
+ XLogSegNo EndSeg;
+
SpinLockAcquire(&XLogCtl->info_lck);
/* advance global request to include new block(s) */
if (XLogCtl->LogwrtRqst.Write < EndPos)
@@ -1165,12 +1197,38 @@ XLogInsertRecord(XLogRecData *rdata,
/* update local result copy while I have the chance */
LogwrtResult = XLogCtl->LogwrtResult;
SpinLockRelease(&XLogCtl->info_lck);
+
+ /*
+ * If we crossed the segment boundary, record it. This is used to
+ * ensure that segments are not marked ready for archival before the
+ * entire record has been flushed to disk.
+ *
+ * Note that we do not use XLByteToPrevSeg() for determining the
+ * ending segment. This is done so that a record that fits perfectly
+ * into the end of the segment is marked ready for archival as soon as
+ * the flushed pointer jumps to the next segment.
+ */
+ XLByteToSeg(StartPos, StartSeg, wal_segment_size);
+ XLByteToSeg(EndPos, EndSeg, wal_segment_size);
+
+ if (StartSeg != EndSeg && XLogArchivingActive())
+ {
+ RegisterSegmentBoundaryEntry(EndSeg, EndPos);
+
+ /*
+ * There's a chance that the record was already flushed to disk
+ * and we missed marking segments as ready for archive. If this
+ * happens, we nudge the WALWriter, which will take care of
+ * notifying segments as needed.
+ */
+ if (LogwrtResult.Flush > EndPos && ProcGlobal->walwriterLatch)
+ SetLatch(ProcGlobal->walwriterLatch);
+ }
}
/*
* If this was an XLOG_SWITCH record, flush the record and the empty
- * padding space that fills the rest of the segment, and perform
- * end-of-segment actions (eg, notifying archiver).
+ * padding space that fills the rest of the segment.
*/
if (isLogSwitch)
{
@@ -1264,6 +1322,192 @@ XLogInsertRecord(XLogRecData *rdata,
return EndPos;
}
+/*
+ * RegisterSegmentBoundaryEntry
+ *
+ * This enters a new entry into the segment boundary map, which is used for
+ * determing when it is safe to mark a segment as ready for archival. An entry
+ * with the given key (the segment number) must not already exist in the map.
+ * Also, the caller is responsible for ensuring that XLByteToSeg() would return
+ * the same segment number for the given record pointer.
+ */
+static void
+RegisterSegmentBoundaryEntry(XLogSegNo seg, XLogRecPtr pos)
+{
+ SegmentBoundaryEntry *entry;
+ bool found;
+
+ LWLockAcquire(SegmentBoundaryLock, LW_EXCLUSIVE);
+
+ entry = (SegmentBoundaryEntry *) hash_search(SegmentBoundaryMap,
+ (void *) &seg, HASH_ENTER,
+ &found);
+ if (found)
+ elog(ERROR, "entry for segment already exists");
+
+ entry->endpos = pos;
+
+ LWLockRelease(SegmentBoundaryLock);
+}
+
+/*
+ * NotifySegmentsReadyForArchive
+ *
+ * This function marks segments as ready for archival, given that it is safe to
+ * do so. It is safe to call this function repeatedly, even if nothing has
+ * changed since the last time it was called.
+ */
+void
+NotifySegmentsReadyForArchive(XLogRecPtr flushRecPtr)
+{
+ XLogSegNo flushed_seg;
+ XLogSegNo latest_boundary_seg;
+ XLogSegNo last_notified;
+
+ /*
+ * We first do a quick sanity check to see if we can bail out without
+ * taking the SegmentBoundaryLock at all. It is expected that this
+ * function will run frequently and that it will need to do nothing the
+ * vast majority of the time.
+ *
+ * Specifically, we bail out if we've already marked the segment prior to
+ * the segment that contains flushRecPtr as ready for archival. We
+ * intentionally use XLByteToSeg() instead of XLByteToPrevSeg() so that we
+ * don't skip notifying when a record fits perfectly into the end of a
+ * segment. (flushRecPtr should point to the first byte of the record
+ * _after_ the one that is known to be flushed to disk.)
+ */
+ last_notified = GetLastNotifiedSegment();
+ XLByteToSeg(flushRecPtr, flushed_seg, wal_segment_size);
+ if (last_notified >= flushed_seg - 1)
+ return;
+
+ LWLockAcquire(SegmentBoundaryLock, LW_EXCLUSIVE);
+
+ /* Reobtain lastNotifiedSeg in case someone else changed it. */
+ last_notified = GetLastNotifiedSegment();
+
+ /* Retrieve the latest segment boundary to use for notifying segments. */
+ if (GetLatestSegmentBoundary(last_notified, flushRecPtr, &latest_boundary_seg))
+ {
+ /*
+ * Update shared memory and discard segment boundaries that are no
+ * longer needed.
+ *
+ * It is safe to update shared memory before we attempt to create the
+ * .ready files. If our calls to XLogArchiveNotifySeg() fail,
+ * RemoveOldXlogFiles() will retry it as needed.
+ */
+ SetLastNotifiedSegment(latest_boundary_seg - 1);
+ RemoveSegmentBoundariesUpTo(latest_boundary_seg);
+
+ LWLockRelease(SegmentBoundaryLock);
+
+ /*
+ * Notify archiver about segments that are ready for archival (by
+ * creating the corresponding .ready files).
+ */
+ for (XLogSegNo seg = last_notified + 1; seg < latest_boundary_seg; seg++)
+ XLogArchiveNotifySeg(seg, false);
+
+ PgArchWakeup();
+ }
+ else
+ LWLockRelease(SegmentBoundaryLock);
+}
+
+/*
+ * GetLatestSegmentBoundary
+ *
+ * This function finds the latest segment boundary in SegmentBoundaryMap that is
+ * less than or equal to the given "flushed" pointer and beyond the last
+ * notified segment. If such a segment is found, latest_boundary_seg is
+ * populated and true is returned. Otherwise, false is returned.
+ */
+static bool
+GetLatestSegmentBoundary(XLogSegNo last_notified, XLogRecPtr flushed,
+ XLogSegNo *latest_boundary_seg)
+{
+ XLogSegNo flushed_seg;
+ XLogSegNo seg;
+
+ Assert(LWLockHeldByMe(SegmentBoundaryLock));
+ Assert(latest_boundary_seg != NULL);
+
+ XLByteToSeg(flushed, flushed_seg, wal_segment_size);
+
+ for (seg = flushed_seg; seg > last_notified; seg--)
+ {
+ SegmentBoundaryEntry *entry;
+
+ entry = (SegmentBoundaryEntry *) hash_search(SegmentBoundaryMap,
+ (void *) &seg, HASH_FIND,
+ NULL);
+
+ if (entry != NULL && flushed >= entry->endpos)
+ {
+ *latest_boundary_seg = entry->seg;
+ return true;
+ }
+ }
+
+ return false;
+}
+
+/*
+ * RemoveSegmentBoundariesUpTo
+ *
+ * This function removes all entries in the SegmentBoundaryMap with segment
+ * numbers up to and including seg.
+ */
+static void
+RemoveSegmentBoundariesUpTo(XLogSegNo seg)
+{
+ SegmentBoundaryEntry *entry;
+ HASH_SEQ_STATUS status;
+
+ Assert(LWLockHeldByMeInMode(SegmentBoundaryLock, LW_EXCLUSIVE));
+
+ hash_seq_init(&status, SegmentBoundaryMap);
+
+ while ((entry = (SegmentBoundaryEntry *) hash_seq_search(&status)) != NULL)
+ {
+ if (entry->seg <= seg)
+ (void) hash_search(SegmentBoundaryMap, (void *) &entry->seg,
+ HASH_REMOVE, NULL);
+ }
+}
+
+/*
+ * GetLastNotifiedSegment
+ *
+ * Retrieves last notified segment from shared memory.
+ */
+XLogSegNo
+GetLastNotifiedSegment(void)
+{
+ XLogSegNo seg;
+
+ SpinLockAcquire(&XLogCtl->info_lck);
+ seg = XLogCtl->lastNotifiedSeg;
+ SpinLockRelease(&XLogCtl->info_lck);
+
+ return seg;
+}
+
+/*
+ * SetLastNotifiedSegment
+ *
+ * Sets last notified segment in shared memory.
+ */
+static void
+SetLastNotifiedSegment(XLogSegNo seg)
+{
+ SpinLockAcquire(&XLogCtl->info_lck);
+ XLogCtl->lastNotifiedSeg = seg;
+ SpinLockRelease(&XLogCtl->info_lck);
+}
+
/*
* Reserves the right amount of space for a record of given size from the WAL.
* *StartPos is set to the beginning of the reserved section, *EndPos to
@@ -2421,6 +2665,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
/* We should always be inside a critical section here */
Assert(CritSectionCount > 0);
+ Assert(LWLockHeldByMe(WALWriteLock));
/*
* Update local LogwrtResult (caller probably did this already, but...)
@@ -2586,11 +2831,13 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
* later. Doing it here ensures that one and only one backend will
* perform this fsync.
*
- * This is also the right place to notify the Archiver that the
- * segment is ready to copy to archival storage, and to update the
- * timer for archive_timeout, and to signal for a checkpoint if
- * too many logfile segments have been used since the last
- * checkpoint.
+ * If WAL archiving is active, we attempt to notify the archiver
+ * of any segments that are now ready for archival.
+ *
+ * This is also the right place to update the timer for
+ * archive_timeout and to signal for a checkpoint if too many
+ * logfile segments have been used since the last checkpoint.
+ *
*/
if (finishing_seg)
{
@@ -2602,7 +2849,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
LogwrtResult.Flush = LogwrtResult.Write; /* end of page */
if (XLogArchivingActive())
- XLogArchiveNotifySeg(openLogSegNo);
+ NotifySegmentsReadyForArchive(LogwrtResult.Flush);
XLogCtl->lastSegSwitchTime = (pg_time_t) time(NULL);
XLogCtl->lastSegSwitchLSN = LogwrtResult.Flush;
@@ -2690,6 +2937,9 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
XLogCtl->LogwrtRqst.Flush = LogwrtResult.Flush;
SpinLockRelease(&XLogCtl->info_lck);
}
+
+ if (XLogArchivingActive())
+ NotifySegmentsReadyForArchive(LogwrtResult.Flush);
}
/*
@@ -5117,6 +5367,9 @@ XLOGShmemSize(void)
/* and the buffers themselves */
size = add_size(size, mul_size(XLOG_BLCKSZ, XLOGbuffers));
+ /* stuff for marking segments as ready for archival */
+ size = add_size(size, hash_estimate_size(16, sizeof(SegmentBoundaryEntry)));
+
/*
* Note: we don't count ControlFileData, it comes out of the "slop factor"
* added by CreateSharedMemoryAndSemaphores. This lets us use this
@@ -5134,6 +5387,7 @@ XLOGShmemInit(void)
char *allocptr;
int i;
ControlFileData *localControlFile;
+ HASHCTL info;
#ifdef WAL_DEBUG
@@ -5227,12 +5481,20 @@ XLOGShmemInit(void)
XLogCtl->InstallXLogFileSegmentActive = false;
XLogCtl->SharedPromoteIsTriggered = false;
XLogCtl->WalWriterSleeping = false;
+ XLogCtl->lastNotifiedSeg = MaxXLogSegNo;
SpinLockInit(&XLogCtl->Insert.insertpos_lck);
SpinLockInit(&XLogCtl->info_lck);
SpinLockInit(&XLogCtl->ulsn_lck);
InitSharedLatch(&XLogCtl->recoveryWakeupLatch);
ConditionVariableInit(&XLogCtl->recoveryNotPausedCV);
+
+ /* Initialize stuff for marking segments as ready for archival. */
+ memset(&info, 0, sizeof(info));
+ info.keysize = sizeof(XLogSegNo);
+ info.entrysize = sizeof(SegmentBoundaryEntry);
+ SegmentBoundaryMap = ShmemInitHash("Segment Boundary Table", 16, 16, &info,
+ HASH_ELEM | HASH_BLOBS);
}
/*
@@ -7873,6 +8135,17 @@ StartupXLOG(void)
XLogCtl->LogwrtRqst.Write = EndOfLog;
XLogCtl->LogwrtRqst.Flush = EndOfLog;
+ /*
+ * Initialize XLogCtl->lastNotifiedSeg to the previous WAL file.
+ */
+ if (XLogArchivingActive())
+ {
+ XLogSegNo EndOfLogSeg;
+
+ XLByteToSeg(EndOfLog, EndOfLogSeg, wal_segment_size);
+ SetLastNotifiedSegment(EndOfLogSeg - 1);
+ }
+
/*
* Update full_page_writes in shared memory and write an XLOG_FPW_CHANGE
* record before resource manager writes cleanup WAL records or checkpoint
@@ -8000,7 +8273,7 @@ StartupXLOG(void)
XLogArchiveCleanup(partialfname);
durable_rename(origpath, partialpath, ERROR);
- XLogArchiveNotify(partialfname);
+ XLogArchiveNotify(partialfname, true);
}
}
}
diff --git a/src/backend/access/transam/xlogarchive.c b/src/backend/access/transam/xlogarchive.c
index 26b023e754..b9c19b2085 100644
--- a/src/backend/access/transam/xlogarchive.c
+++ b/src/backend/access/transam/xlogarchive.c
@@ -433,7 +433,7 @@ KeepFileRestoredFromArchive(const char *path, const char *xlogfname)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(xlogfname);
else
- XLogArchiveNotify(xlogfname);
+ XLogArchiveNotify(xlogfname, true);
/*
* If the existing file was replaced, since walsenders might have it open,
@@ -462,9 +462,12 @@ KeepFileRestoredFromArchive(const char *path, const char *xlogfname)
* by the archiver, e.g. we write 0000000100000001000000C6.ready
* and the archiver then knows to archive XLOGDIR/0000000100000001000000C6,
* then when complete, rename it to 0000000100000001000000C6.done
+ *
+ * Optionally, nudge the archiver process so that it'll notice the file we
+ * create.
*/
void
-XLogArchiveNotify(const char *xlog)
+XLogArchiveNotify(const char *xlog, bool nudge)
{
char archiveStatusPath[MAXPGPATH];
FILE *fd;
@@ -489,8 +492,8 @@ XLogArchiveNotify(const char *xlog)
return;
}
- /* Notify archiver that it's got something to do */
- if (IsUnderPostmaster)
+ /* If caller requested, let archiver know it's got work to do */
+ if (nudge)
PgArchWakeup();
}
@@ -498,12 +501,12 @@ XLogArchiveNotify(const char *xlog)
* Convenience routine to notify using segment number representation of filename
*/
void
-XLogArchiveNotifySeg(XLogSegNo segno)
+XLogArchiveNotifySeg(XLogSegNo segno, bool nudge)
{
char xlog[MAXFNAMELEN];
XLogFileName(xlog, ThisTimeLineID, segno, wal_segment_size);
- XLogArchiveNotify(xlog);
+ XLogArchiveNotify(xlog, nudge);
}
/*
@@ -608,7 +611,7 @@ XLogArchiveCheckDone(const char *xlog)
return true;
/* Retry creation of the .ready file */
- XLogArchiveNotify(xlog);
+ XLogArchiveNotify(xlog, true);
return false;
}
diff --git a/src/backend/postmaster/walwriter.c b/src/backend/postmaster/walwriter.c
index 626fae8454..6a1e16edc2 100644
--- a/src/backend/postmaster/walwriter.c
+++ b/src/backend/postmaster/walwriter.c
@@ -248,6 +248,13 @@ WalWriterMain(void)
/* Process any signals received recently */
HandleWalWriterInterrupts();
+ /*
+ * Notify the archiver of any WAL segments that are ready. We do this
+ * here to handle a race condition where WAL is flushed to disk prior
+ * to registering the segment boundary.
+ */
+ NotifySegmentsReadyForArchive(GetFlushRecPtr());
+
/*
* Do what we're here for; then, if XLogBackgroundFlush() found useful
* work to do, reset hibernation counter.
diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c
index 9a2bc37fd7..60de3be92c 100644
--- a/src/backend/replication/walreceiver.c
+++ b/src/backend/replication/walreceiver.c
@@ -622,7 +622,7 @@ WalReceiverMain(void)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(xlogfname);
else
- XLogArchiveNotify(xlogfname);
+ XLogArchiveNotify(xlogfname, true);
}
recvFile = -1;
@@ -760,7 +760,7 @@ WalRcvFetchTimeLineHistoryFiles(TimeLineID first, TimeLineID last)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(fname);
else
- XLogArchiveNotify(fname);
+ XLogArchiveNotify(fname, true);
pfree(fname);
pfree(content);
@@ -915,7 +915,7 @@ XLogWalRcvWrite(char *buf, Size nbytes, XLogRecPtr recptr)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(xlogfname);
else
- XLogArchiveNotify(xlogfname);
+ XLogArchiveNotify(xlogfname, true);
}
recvFile = -1;
diff --git a/src/backend/storage/lmgr/lwlocknames.txt b/src/backend/storage/lmgr/lwlocknames.txt
index 6c7cf6c295..14a742d655 100644
--- a/src/backend/storage/lmgr/lwlocknames.txt
+++ b/src/backend/storage/lmgr/lwlocknames.txt
@@ -53,3 +53,4 @@ XactTruncationLock 44
# 45 was XactTruncationLock until removal of BackendRandomLock
WrapLimitsVacuumLock 46
NotifyQueueTailLock 47
+SegmentBoundaryLock 48
diff --git a/src/include/access/xlog.h b/src/include/access/xlog.h
index 0a8ede700d..6b6ae81c2d 100644
--- a/src/include/access/xlog.h
+++ b/src/include/access/xlog.h
@@ -315,6 +315,7 @@ extern XLogRecPtr GetInsertRecPtr(void);
extern XLogRecPtr GetFlushRecPtr(void);
extern XLogRecPtr GetLastImportantRecPtr(void);
extern void RemovePromoteSignalFiles(void);
+extern void NotifySegmentsReadyForArchive(XLogRecPtr flushRecPtr);
extern bool PromoteIsTriggered(void);
extern bool CheckPromoteSignal(void);
diff --git a/src/include/access/xlogarchive.h b/src/include/access/xlogarchive.h
index 3edd1a976c..935b4cb02d 100644
--- a/src/include/access/xlogarchive.h
+++ b/src/include/access/xlogarchive.h
@@ -23,8 +23,8 @@ extern bool RestoreArchivedFile(char *path, const char *xlogfname,
extern void ExecuteRecoveryCommand(const char *command, const char *commandName,
bool failOnSignal);
extern void KeepFileRestoredFromArchive(const char *path, const char *xlogfname);
-extern void XLogArchiveNotify(const char *xlog);
-extern void XLogArchiveNotifySeg(XLogSegNo segno);
+extern void XLogArchiveNotify(const char *xlog, bool nudge);
+extern void XLogArchiveNotifySeg(XLogSegNo segno, bool nudge);
extern void XLogArchiveForceDone(const char *xlog);
extern bool XLogArchiveCheckDone(const char *xlog);
extern bool XLogArchiveIsBusy(const char *xlog);
diff --git a/src/include/access/xlogdefs.h b/src/include/access/xlogdefs.h
index 60348d1850..9b455e88e3 100644
--- a/src/include/access/xlogdefs.h
+++ b/src/include/access/xlogdefs.h
@@ -46,6 +46,7 @@ typedef uint64 XLogRecPtr;
* XLogSegNo - physical log file sequence number.
*/
typedef uint64 XLogSegNo;
+#define MaxXLogSegNo ((XLogSegNo) 0xFFFFFFFFFFFFFFFF)
/*
* TimeLineID (TLI) - identifies different database histories to prevent
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 37cf4b2f76..79694b049e 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2334,6 +2334,7 @@ SecBufferDesc
SecLabelItem
SecLabelStmt
SeenRelsEntry
+SegmentBoundaryEntry
SelectLimit
SelectStmt
Selectivity
--
2.30.2
--aabamu5eazcf6rc5--
^ permalink raw reply [nested|flat] 19+ messages in thread
* [PATCH v12] Avoid creating archive status ".ready" files too early
@ 2021-08-17 03:52 Nathan Bossart <[email protected]>
0 siblings, 0 replies; 19+ messages in thread
From: Nathan Bossart @ 2021-08-17 03:52 UTC (permalink / raw)
WAL records may span multiple segments, but XLogWrite() does not
wait for the entire record to be written out to disk before
creating archive status files. Instead, as soon as the last WAL page of
the segment is written, the archive status file is created, and the
archiver may process it. If PostgreSQL crashes before it is able to
write and flush the rest of the record (in the next WAL segment), the
wrong version of the first segment file lingers in the archive, which
causes operations such as point-in-time restores to fail.
To fix this, keep track of records that span across segments and ensure
that segments are only marked ready-for-archival once such records have
been completely written to disk.
XXX Note that this may add a new LWLock acquisition in the transaction
commit path. (It doesn't happen in all cases.)
Author: Nathan Bossart <[email protected]>
Reviewed-by: Kyotaro Horiguchi <[email protected]>
Reviewed-by: Ryo Matsumura <[email protected]>
Reviewed-by: Andrey Borodin <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
src/backend/access/transam/timeline.c | 2 +-
src/backend/access/transam/xlog.c | 307 ++++++++++++++++++++++-
src/backend/access/transam/xlogarchive.c | 17 +-
src/backend/postmaster/walwriter.c | 7 +
src/backend/replication/walreceiver.c | 6 +-
src/backend/storage/lmgr/lwlocknames.txt | 1 +
src/include/access/xlog.h | 1 +
src/include/access/xlogarchive.h | 4 +-
src/include/access/xlogdefs.h | 1 +
src/tools/pgindent/typedefs.list | 1 +
10 files changed, 323 insertions(+), 24 deletions(-)
diff --git a/src/backend/access/transam/timeline.c b/src/backend/access/transam/timeline.c
index 8d0903c175..acd5c2431d 100644
--- a/src/backend/access/transam/timeline.c
+++ b/src/backend/access/transam/timeline.c
@@ -452,7 +452,7 @@ writeTimeLineHistory(TimeLineID newTLI, TimeLineID parentTLI,
if (XLogArchivingActive())
{
TLHistoryFileName(histfname, newTLI);
- XLogArchiveNotify(histfname);
+ XLogArchiveNotify(histfname, true);
}
}
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e51a7a749d..01da4ab7b8 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -512,6 +512,16 @@ typedef enum ExclusiveBackupState
*/
static SessionBackupState sessionBackupState = SESSION_BACKUP_NONE;
+/*
+ * Entries for SegmentBoundaryMap. Each such entry represents a WAL
+ * record that ends in endpos and crosses the WAL segment boundary.
+ */
+typedef struct SegmentBoundaryEntry
+{
+ XLogSegNo seg; /* must be first */
+ XLogRecPtr endpos;
+} SegmentBoundaryEntry;
+
/*
* Shared state data for WAL insertion.
*/
@@ -723,6 +733,12 @@ typedef struct XLogCtlData
*/
XLogRecPtr lastFpwDisableRecPtr;
+ /*
+ * The last segment we've marked ready for archival. Protected by
+ * info_lck.
+ */
+ XLogSegNo lastNotifiedSeg;
+
slock_t info_lck; /* locks shared variables shown above */
} XLogCtlData;
@@ -736,6 +752,23 @@ static WALInsertLockPadded *WALInsertLocks = NULL;
*/
static ControlFileData *ControlFile = NULL;
+/*
+ * Segment boundary map.
+ *
+ * This hash table tracks WAL records that cross WAL segment boundaries.
+ * For WAL archiving it is critical not to mark a segment as archivable
+ * until both halves of the record are flushed: otherwise if we crash
+ * after archiving the segment containing the first half and before
+ * flushing the second half of the record, then after recovery the
+ * primary would overwrite the first half with new data, making the
+ * segment archived prior to the crash a bogus copy.
+ *
+ * XXX note that an equivalent problem exists with streaming replication.
+ *
+ * Protected by SegmentBoundaryLock.
+ */
+static HTAB *SegmentBoundaryMap = NULL;
+
/*
* Calculate the amount of space left on the page after 'endptr'. Beware
* multiple evaluation!
@@ -920,6 +953,13 @@ static void RemoveXlogFile(const char *segname, XLogSegNo recycleSegNo,
XLogSegNo *endlogSegNo);
static void UpdateLastRemovedPtr(char *filename);
static void ValidateXLOGDirectoryStructure(void);
+static void RegisterSegmentBoundary(XLogSegNo seg, XLogRecPtr pos);
+static XLogSegNo GetLastNotifiedSegment(void);
+static void SetLastNotifiedSegment(XLogSegNo seg);
+static bool GetLatestSegmentBoundary(XLogSegNo last_notified,
+ XLogRecPtr flushed,
+ XLogSegNo *latest_boundary_seg);
+static void RemoveSegmentBoundariesUpTo(XLogSegNo seg);
static void CleanupBackupHistory(void);
static void UpdateMinRecoveryPoint(XLogRecPtr lsn, bool force);
static XLogRecord *ReadRecord(XLogReaderState *xlogreader,
@@ -1154,23 +1194,56 @@ XLogInsertRecord(XLogRecData *rdata,
END_CRIT_SECTION();
/*
- * Update shared LogwrtRqst.Write, if we crossed page boundary.
+ * If we crossed page boundary, update LogwrtRqst.Write; if we crossed
+ * segment boundary, register that and wake up walwriter.
*/
if (StartPos / XLOG_BLCKSZ != EndPos / XLOG_BLCKSZ)
{
+ XLogSegNo StartSeg;
+ XLogSegNo EndSeg;
+
+ XLByteToSeg(StartPos, StartSeg, wal_segment_size);
+ XLByteToSeg(EndPos, EndSeg, wal_segment_size);
+
+ /*
+ * Register our crossing the segment boundary if that occurred.
+ *
+ * Note that we did not use XLByteToPrevSeg() for determining the
+ * ending segment. This is so that a record that fits perfectly into
+ * the end of the segment is marked ready for archival as soon as the
+ * flushed pointer jumps to the next segment.
+ */
+ if (StartSeg != EndSeg && XLogArchivingActive())
+ RegisterSegmentBoundary(EndSeg, EndPos);
+
+ /*
+ * Advance LogwrtRqst.Write so that it includes new block(s).
+ *
+ * We do this after registering the segment boundary so that the
+ * comparison with the flushed pointer below can use the latest value
+ * known globally.
+ */
SpinLockAcquire(&XLogCtl->info_lck);
- /* advance global request to include new block(s) */
if (XLogCtl->LogwrtRqst.Write < EndPos)
XLogCtl->LogwrtRqst.Write = EndPos;
/* update local result copy while I have the chance */
LogwrtResult = XLogCtl->LogwrtResult;
SpinLockRelease(&XLogCtl->info_lck);
+
+ /*
+ * There's a chance that the record was already flushed to disk and we
+ * missed marking segments as ready for archive. If this happens, we
+ * nudge the WALWriter, which will take care of notifying segments as
+ * needed.
+ */
+ if (StartSeg != EndSeg && XLogArchivingActive() &&
+ LogwrtResult.Flush >= EndPos && ProcGlobal->walwriterLatch)
+ SetLatch(ProcGlobal->walwriterLatch);
}
/*
* If this was an XLOG_SWITCH record, flush the record and the empty
- * padding space that fills the rest of the segment, and perform
- * end-of-segment actions (eg, notifying archiver).
+ * padding space that fills the rest of the segment.
*/
if (isLogSwitch)
{
@@ -2421,6 +2494,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
/* We should always be inside a critical section here */
Assert(CritSectionCount > 0);
+ Assert(LWLockHeldByMe(WALWriteLock));
/*
* Update local LogwrtResult (caller probably did this already, but...)
@@ -2586,11 +2660,13 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
* later. Doing it here ensures that one and only one backend will
* perform this fsync.
*
- * This is also the right place to notify the Archiver that the
- * segment is ready to copy to archival storage, and to update the
- * timer for archive_timeout, and to signal for a checkpoint if
- * too many logfile segments have been used since the last
- * checkpoint.
+ * If WAL archiving is active, we attempt to notify the archiver
+ * of any segments that are now ready for archival.
+ *
+ * This is also the right place to update the timer for
+ * archive_timeout and to signal for a checkpoint if too many
+ * logfile segments have been used since the last checkpoint.
+ *
*/
if (finishing_seg)
{
@@ -2602,7 +2678,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
LogwrtResult.Flush = LogwrtResult.Write; /* end of page */
if (XLogArchivingActive())
- XLogArchiveNotifySeg(openLogSegNo);
+ NotifySegmentsReadyForArchive(LogwrtResult.Flush);
XLogCtl->lastSegSwitchTime = (pg_time_t) time(NULL);
XLogCtl->lastSegSwitchLSN = LogwrtResult.Flush;
@@ -2690,6 +2766,9 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
XLogCtl->LogwrtRqst.Flush = LogwrtResult.Flush;
SpinLockRelease(&XLogCtl->info_lck);
}
+
+ if (XLogArchivingActive())
+ NotifySegmentsReadyForArchive(LogwrtResult.Flush);
}
/*
@@ -4328,6 +4407,189 @@ ValidateXLOGDirectoryStructure(void)
}
}
+/*
+ * RegisterSegmentBoundary
+ * Enter a new entry into the segment boundary map.
+ *
+ * An entry with the given key (the segment number) must not already exist in
+ * the map.
+ */
+static void
+RegisterSegmentBoundary(XLogSegNo seg, XLogRecPtr pos)
+{
+ SegmentBoundaryEntry *entry;
+ XLogSegNo segno PG_USED_FOR_ASSERTS_ONLY;
+ bool found;
+
+ /* verify caller computed segment number correctly */
+ AssertArg((XLByteToSeg(pos, segno, wal_segment_size), segno == seg));
+
+ LWLockAcquire(SegmentBoundaryLock, LW_EXCLUSIVE);
+
+ entry = (SegmentBoundaryEntry *) hash_search(SegmentBoundaryMap,
+ (void *) &seg, HASH_ENTER,
+ &found);
+ if (found)
+ elog(ERROR, "entry for segment already exists");
+
+ entry->endpos = pos;
+
+ LWLockRelease(SegmentBoundaryLock);
+}
+
+/*
+ * GetLatestSegmentBoundary
+ *
+ * Find the latest segment boundary in SegmentBoundaryMap that is less
+ * than or equal to the given "flushed" pointer and beyond the last
+ * notified segment. If such a segment is found, latest_boundary_seg
+ * is populated and true is returned. Otherwise, false is returned.
+ */
+static bool
+GetLatestSegmentBoundary(XLogSegNo last_notified, XLogRecPtr flushed,
+ XLogSegNo *latest_boundary_seg)
+{
+ XLogSegNo flushed_seg;
+ XLogSegNo seg;
+
+ Assert(LWLockHeldByMe(SegmentBoundaryLock));
+ Assert(latest_boundary_seg != NULL);
+
+ XLByteToSeg(flushed, flushed_seg, wal_segment_size);
+
+ for (seg = flushed_seg; seg > last_notified; seg--)
+ {
+ SegmentBoundaryEntry *entry;
+
+ entry = (SegmentBoundaryEntry *) hash_search(SegmentBoundaryMap,
+ (void *) &seg, HASH_FIND,
+ NULL);
+
+ if (entry != NULL && flushed >= entry->endpos)
+ {
+ *latest_boundary_seg = entry->seg;
+ return true;
+ }
+ }
+
+ return false;
+}
+
+/*
+ * RemoveSegmentBoundariesUpTo
+ *
+ * Remove all entries in the SegmentBoundaryMap with segment numbers
+ * up to and including seg.
+ */
+static void
+RemoveSegmentBoundariesUpTo(XLogSegNo seg)
+{
+ SegmentBoundaryEntry *entry;
+ HASH_SEQ_STATUS status;
+
+ Assert(LWLockHeldByMeInMode(SegmentBoundaryLock, LW_EXCLUSIVE));
+
+ hash_seq_init(&status, SegmentBoundaryMap);
+
+ while ((entry = (SegmentBoundaryEntry *) hash_seq_search(&status)) != NULL)
+ {
+ if (entry->seg <= seg)
+ (void) hash_search(SegmentBoundaryMap, (void *) &entry->seg,
+ HASH_REMOVE, NULL);
+ }
+}
+
+/*
+ * NotifySegmentsReadyForArchive
+ *
+ * Mark segments as ready for archival, given that it is safe to do so.
+ * This function is idempotent.
+ */
+void
+NotifySegmentsReadyForArchive(XLogRecPtr flushRecPtr)
+{
+ XLogSegNo flushed_seg;
+ XLogSegNo latest_boundary_seg;
+ XLogSegNo last_notified;
+
+ /*
+ * We first do a quick sanity check to see if we can bail out without
+ * taking the SegmentBoundaryLock at all. It is expected that this
+ * function will run frequently and that it will need to do nothing the
+ * vast majority of the time.
+ *
+ * Specifically, we bail out if we've already marked the segment prior to
+ * the segment that contains flushRecPtr as ready for archival. We
+ * intentionally use XLByteToSeg() instead of XLByteToPrevSeg() so that we
+ * don't skip notifying when a record fits perfectly into the end of a
+ * segment. (flushRecPtr should point to the first byte of the record
+ * _after_ the one that is known to be flushed to disk.)
+ */
+ last_notified = GetLastNotifiedSegment();
+ XLByteToSeg(flushRecPtr, flushed_seg, wal_segment_size);
+ if (last_notified >= flushed_seg - 1)
+ return;
+
+ LWLockAcquire(SegmentBoundaryLock, LW_EXCLUSIVE);
+
+ /* Reobtain lastNotifiedSeg in case someone else changed it. */
+ last_notified = GetLastNotifiedSegment();
+
+ /* Retrieve the latest segment boundary to use for notifying segments. */
+ if (GetLatestSegmentBoundary(last_notified, flushRecPtr, &latest_boundary_seg))
+ {
+ /*
+ * Update shared memory and discard segment boundaries that are no
+ * longer needed.
+ *
+ * It is safe to update shared memory before we attempt to create the
+ * .ready files. If our calls to XLogArchiveNotifySeg() fail,
+ * RemoveOldXlogFiles() will retry it as needed.
+ */
+ SetLastNotifiedSegment(latest_boundary_seg - 1);
+ RemoveSegmentBoundariesUpTo(latest_boundary_seg);
+
+ LWLockRelease(SegmentBoundaryLock);
+
+ /*
+ * Notify archiver about segments that are ready for archival (by
+ * creating the corresponding .ready files).
+ */
+ for (XLogSegNo seg = last_notified + 1; seg < latest_boundary_seg; seg++)
+ XLogArchiveNotifySeg(seg, false);
+
+ PgArchWakeup();
+ }
+ else
+ LWLockRelease(SegmentBoundaryLock);
+}
+
+/*
+ * GetLastNotifiedSegment
+ */
+XLogSegNo
+GetLastNotifiedSegment(void)
+{
+ XLogSegNo seg;
+
+ SpinLockAcquire(&XLogCtl->info_lck);
+ seg = XLogCtl->lastNotifiedSeg;
+ SpinLockRelease(&XLogCtl->info_lck);
+
+ return seg;
+}
+
+/*
+ * SetLastNotifiedSegment
+ */
+static void
+SetLastNotifiedSegment(XLogSegNo seg)
+{
+ SpinLockAcquire(&XLogCtl->info_lck);
+ XLogCtl->lastNotifiedSeg = seg;
+ SpinLockRelease(&XLogCtl->info_lck);
+}
+
/*
* Remove previous backup history files. This also retries creation of
* .ready files for any backup history files for which XLogArchiveNotify
@@ -5117,6 +5379,9 @@ XLOGShmemSize(void)
/* and the buffers themselves */
size = add_size(size, mul_size(XLOG_BLCKSZ, XLOGbuffers));
+ /* hash table of segment-crossing WAL records */
+ size = add_size(size, hash_estimate_size(16, sizeof(SegmentBoundaryEntry)));
+
/*
* Note: we don't count ControlFileData, it comes out of the "slop factor"
* added by CreateSharedMemoryAndSemaphores. This lets us use this
@@ -5134,6 +5399,7 @@ XLOGShmemInit(void)
char *allocptr;
int i;
ControlFileData *localControlFile;
+ HASHCTL info;
#ifdef WAL_DEBUG
@@ -5227,12 +5493,20 @@ XLOGShmemInit(void)
XLogCtl->InstallXLogFileSegmentActive = false;
XLogCtl->SharedPromoteIsTriggered = false;
XLogCtl->WalWriterSleeping = false;
+ XLogCtl->lastNotifiedSeg = MaxXLogSegNo;
SpinLockInit(&XLogCtl->Insert.insertpos_lck);
SpinLockInit(&XLogCtl->info_lck);
SpinLockInit(&XLogCtl->ulsn_lck);
InitSharedLatch(&XLogCtl->recoveryWakeupLatch);
ConditionVariableInit(&XLogCtl->recoveryNotPausedCV);
+
+ /* Initialize stuff for marking segments as ready for archival. */
+ memset(&info, 0, sizeof(info));
+ info.keysize = sizeof(XLogSegNo);
+ info.entrysize = sizeof(SegmentBoundaryEntry);
+ SegmentBoundaryMap = ShmemInitHash("Segment Boundary Table", 16, 16, &info,
+ HASH_ELEM | HASH_BLOBS);
}
/*
@@ -7873,6 +8147,17 @@ StartupXLOG(void)
XLogCtl->LogwrtRqst.Write = EndOfLog;
XLogCtl->LogwrtRqst.Flush = EndOfLog;
+ /*
+ * Initialize XLogCtl->lastNotifiedSeg to the previous WAL file.
+ */
+ if (XLogArchivingActive())
+ {
+ XLogSegNo EndOfLogSeg;
+
+ XLByteToSeg(EndOfLog, EndOfLogSeg, wal_segment_size);
+ SetLastNotifiedSegment(EndOfLogSeg - 1);
+ }
+
/*
* Update full_page_writes in shared memory and write an XLOG_FPW_CHANGE
* record before resource manager writes cleanup WAL records or checkpoint
@@ -8000,7 +8285,7 @@ StartupXLOG(void)
XLogArchiveCleanup(partialfname);
durable_rename(origpath, partialpath, ERROR);
- XLogArchiveNotify(partialfname);
+ XLogArchiveNotify(partialfname, true);
}
}
}
diff --git a/src/backend/access/transam/xlogarchive.c b/src/backend/access/transam/xlogarchive.c
index 26b023e754..b9c19b2085 100644
--- a/src/backend/access/transam/xlogarchive.c
+++ b/src/backend/access/transam/xlogarchive.c
@@ -433,7 +433,7 @@ KeepFileRestoredFromArchive(const char *path, const char *xlogfname)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(xlogfname);
else
- XLogArchiveNotify(xlogfname);
+ XLogArchiveNotify(xlogfname, true);
/*
* If the existing file was replaced, since walsenders might have it open,
@@ -462,9 +462,12 @@ KeepFileRestoredFromArchive(const char *path, const char *xlogfname)
* by the archiver, e.g. we write 0000000100000001000000C6.ready
* and the archiver then knows to archive XLOGDIR/0000000100000001000000C6,
* then when complete, rename it to 0000000100000001000000C6.done
+ *
+ * Optionally, nudge the archiver process so that it'll notice the file we
+ * create.
*/
void
-XLogArchiveNotify(const char *xlog)
+XLogArchiveNotify(const char *xlog, bool nudge)
{
char archiveStatusPath[MAXPGPATH];
FILE *fd;
@@ -489,8 +492,8 @@ XLogArchiveNotify(const char *xlog)
return;
}
- /* Notify archiver that it's got something to do */
- if (IsUnderPostmaster)
+ /* If caller requested, let archiver know it's got work to do */
+ if (nudge)
PgArchWakeup();
}
@@ -498,12 +501,12 @@ XLogArchiveNotify(const char *xlog)
* Convenience routine to notify using segment number representation of filename
*/
void
-XLogArchiveNotifySeg(XLogSegNo segno)
+XLogArchiveNotifySeg(XLogSegNo segno, bool nudge)
{
char xlog[MAXFNAMELEN];
XLogFileName(xlog, ThisTimeLineID, segno, wal_segment_size);
- XLogArchiveNotify(xlog);
+ XLogArchiveNotify(xlog, nudge);
}
/*
@@ -608,7 +611,7 @@ XLogArchiveCheckDone(const char *xlog)
return true;
/* Retry creation of the .ready file */
- XLogArchiveNotify(xlog);
+ XLogArchiveNotify(xlog, true);
return false;
}
diff --git a/src/backend/postmaster/walwriter.c b/src/backend/postmaster/walwriter.c
index 626fae8454..6a1e16edc2 100644
--- a/src/backend/postmaster/walwriter.c
+++ b/src/backend/postmaster/walwriter.c
@@ -248,6 +248,13 @@ WalWriterMain(void)
/* Process any signals received recently */
HandleWalWriterInterrupts();
+ /*
+ * Notify the archiver of any WAL segments that are ready. We do this
+ * here to handle a race condition where WAL is flushed to disk prior
+ * to registering the segment boundary.
+ */
+ NotifySegmentsReadyForArchive(GetFlushRecPtr());
+
/*
* Do what we're here for; then, if XLogBackgroundFlush() found useful
* work to do, reset hibernation counter.
diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c
index 9a2bc37fd7..60de3be92c 100644
--- a/src/backend/replication/walreceiver.c
+++ b/src/backend/replication/walreceiver.c
@@ -622,7 +622,7 @@ WalReceiverMain(void)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(xlogfname);
else
- XLogArchiveNotify(xlogfname);
+ XLogArchiveNotify(xlogfname, true);
}
recvFile = -1;
@@ -760,7 +760,7 @@ WalRcvFetchTimeLineHistoryFiles(TimeLineID first, TimeLineID last)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(fname);
else
- XLogArchiveNotify(fname);
+ XLogArchiveNotify(fname, true);
pfree(fname);
pfree(content);
@@ -915,7 +915,7 @@ XLogWalRcvWrite(char *buf, Size nbytes, XLogRecPtr recptr)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(xlogfname);
else
- XLogArchiveNotify(xlogfname);
+ XLogArchiveNotify(xlogfname, true);
}
recvFile = -1;
diff --git a/src/backend/storage/lmgr/lwlocknames.txt b/src/backend/storage/lmgr/lwlocknames.txt
index 6c7cf6c295..14a742d655 100644
--- a/src/backend/storage/lmgr/lwlocknames.txt
+++ b/src/backend/storage/lmgr/lwlocknames.txt
@@ -53,3 +53,4 @@ XactTruncationLock 44
# 45 was XactTruncationLock until removal of BackendRandomLock
WrapLimitsVacuumLock 46
NotifyQueueTailLock 47
+SegmentBoundaryLock 48
diff --git a/src/include/access/xlog.h b/src/include/access/xlog.h
index 0a8ede700d..6b6ae81c2d 100644
--- a/src/include/access/xlog.h
+++ b/src/include/access/xlog.h
@@ -315,6 +315,7 @@ extern XLogRecPtr GetInsertRecPtr(void);
extern XLogRecPtr GetFlushRecPtr(void);
extern XLogRecPtr GetLastImportantRecPtr(void);
extern void RemovePromoteSignalFiles(void);
+extern void NotifySegmentsReadyForArchive(XLogRecPtr flushRecPtr);
extern bool PromoteIsTriggered(void);
extern bool CheckPromoteSignal(void);
diff --git a/src/include/access/xlogarchive.h b/src/include/access/xlogarchive.h
index 3edd1a976c..935b4cb02d 100644
--- a/src/include/access/xlogarchive.h
+++ b/src/include/access/xlogarchive.h
@@ -23,8 +23,8 @@ extern bool RestoreArchivedFile(char *path, const char *xlogfname,
extern void ExecuteRecoveryCommand(const char *command, const char *commandName,
bool failOnSignal);
extern void KeepFileRestoredFromArchive(const char *path, const char *xlogfname);
-extern void XLogArchiveNotify(const char *xlog);
-extern void XLogArchiveNotifySeg(XLogSegNo segno);
+extern void XLogArchiveNotify(const char *xlog, bool nudge);
+extern void XLogArchiveNotifySeg(XLogSegNo segno, bool nudge);
extern void XLogArchiveForceDone(const char *xlog);
extern bool XLogArchiveCheckDone(const char *xlog);
extern bool XLogArchiveIsBusy(const char *xlog);
diff --git a/src/include/access/xlogdefs.h b/src/include/access/xlogdefs.h
index 60348d1850..9b455e88e3 100644
--- a/src/include/access/xlogdefs.h
+++ b/src/include/access/xlogdefs.h
@@ -46,6 +46,7 @@ typedef uint64 XLogRecPtr;
* XLogSegNo - physical log file sequence number.
*/
typedef uint64 XLogSegNo;
+#define MaxXLogSegNo ((XLogSegNo) 0xFFFFFFFFFFFFFFFF)
/*
* TimeLineID (TLI) - identifies different database histories to prevent
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index 37cf4b2f76..79694b049e 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2334,6 +2334,7 @@ SecBufferDesc
SecLabelItem
SecLabelStmt
SeenRelsEntry
+SegmentBoundaryEntry
SelectLimit
SelectStmt
Selectivity
--
2.20.1
--ngwxbmustuw56o2o--
^ permalink raw reply [nested|flat] 19+ messages in thread
* [PATCH v14] Avoid creating archive status ".ready" files too early
@ 2021-08-20 19:25 Nathan Bossart <[email protected]>
0 siblings, 0 replies; 19+ messages in thread
From: Nathan Bossart @ 2021-08-20 19:25 UTC (permalink / raw)
WAL records may span multiple segments, but XLogWrite() does not
wait for the entire record to be written out to disk before
creating archive status files. Instead, as soon as the last WAL page of
the segment is written, the archive status file is created, and the
archiver may process it. If PostgreSQL crashes before it is able to
write and flush the rest of the record (in the next WAL segment), the
wrong version of the first segment file lingers in the archive, which
causes operations such as point-in-time restores to fail.
To fix this, keep track of records that span across segments and ensure
that segments are only marked ready-for-archival once such records have
been completely written to disk.
Author: Nathan Bossart <[email protected]>
Reviewed-by: Kyotaro Horiguchi <[email protected]>
Reviewed-by: Ryo Matsumura <[email protected]>
Reviewed-by: Andrey Borodin <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
src/backend/access/transam/timeline.c | 2 +-
src/backend/access/transam/xlog.c | 242 +++++++++++++++++++++--
src/backend/access/transam/xlogarchive.c | 17 +-
src/backend/postmaster/walwriter.c | 7 +
src/backend/replication/walreceiver.c | 6 +-
src/include/access/xlog.h | 1 +
src/include/access/xlogarchive.h | 4 +-
src/include/access/xlogdefs.h | 1 +
8 files changed, 256 insertions(+), 24 deletions(-)
diff --git a/src/backend/access/transam/timeline.c b/src/backend/access/transam/timeline.c
index 8d0903c175..acd5c2431d 100644
--- a/src/backend/access/transam/timeline.c
+++ b/src/backend/access/transam/timeline.c
@@ -452,7 +452,7 @@ writeTimeLineHistory(TimeLineID newTLI, TimeLineID parentTLI,
if (XLogArchivingActive())
{
TLHistoryFileName(histfname, newTLI);
- XLogArchiveNotify(histfname);
+ XLogArchiveNotify(histfname, true);
}
}
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e51a7a749d..95f03adef8 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -724,6 +724,18 @@ typedef struct XLogCtlData
XLogRecPtr lastFpwDisableRecPtr;
slock_t info_lck; /* locks shared variables shown above */
+
+ /*
+ * Variables used to track cross-segment records. Protected by
+ * segtrack_lck.
+ */
+ XLogSegNo lastNotifiedSeg;
+ XLogSegNo earliestSegBoundary;
+ XLogRecPtr earliestSegBoundaryRecPtr;
+ XLogSegNo latestSegBoundary;
+ XLogRecPtr latestSegBoundaryRecPtr;
+
+ slock_t segtrack_lck; /* locks shared variables shown above */
} XLogCtlData;
static XLogCtlData *XLogCtl = NULL;
@@ -920,6 +932,9 @@ static void RemoveXlogFile(const char *segname, XLogSegNo recycleSegNo,
XLogSegNo *endlogSegNo);
static void UpdateLastRemovedPtr(char *filename);
static void ValidateXLOGDirectoryStructure(void);
+static void RegisterSegmentBoundary(XLogSegNo seg, XLogRecPtr pos);
+static bool GetLatestSegmentBoundary(XLogRecPtr flushed, XLogSegNo *latest_boundary_seg);
+static void RemoveSegmentBoundariesUpTo(XLogSegNo seg);
static void CleanupBackupHistory(void);
static void UpdateMinRecoveryPoint(XLogRecPtr lsn, bool force);
static XLogRecord *ReadRecord(XLogReaderState *xlogreader,
@@ -1154,23 +1169,56 @@ XLogInsertRecord(XLogRecData *rdata,
END_CRIT_SECTION();
/*
- * Update shared LogwrtRqst.Write, if we crossed page boundary.
+ * If we crossed page boundary, update LogwrtRqst.Write; if we crossed
+ * segment boundary, register that and wake up walwriter.
*/
if (StartPos / XLOG_BLCKSZ != EndPos / XLOG_BLCKSZ)
{
+ XLogSegNo StartSeg;
+ XLogSegNo EndSeg;
+
+ XLByteToSeg(StartPos, StartSeg, wal_segment_size);
+ XLByteToSeg(EndPos, EndSeg, wal_segment_size);
+
+ /*
+ * Register our crossing the segment boundary if that occurred.
+ *
+ * Note that we did not use XLByteToPrevSeg() for determining the
+ * ending segment. This is so that a record that fits perfectly into
+ * the end of the segment is marked ready for archival as soon as the
+ * flushed pointer jumps to the next segment.
+ */
+ if (StartSeg != EndSeg && XLogArchivingActive())
+ RegisterSegmentBoundary(EndSeg, EndPos);
+
+ /*
+ * Advance LogwrtRqst.Write so that it includes new block(s).
+ *
+ * We do this after registering the segment boundary so that the
+ * comparison with the flushed pointer below can use the latest value
+ * known globally.
+ */
SpinLockAcquire(&XLogCtl->info_lck);
- /* advance global request to include new block(s) */
if (XLogCtl->LogwrtRqst.Write < EndPos)
XLogCtl->LogwrtRqst.Write = EndPos;
/* update local result copy while I have the chance */
LogwrtResult = XLogCtl->LogwrtResult;
SpinLockRelease(&XLogCtl->info_lck);
+
+ /*
+ * There's a chance that the record was already flushed to disk and we
+ * missed marking segments as ready for archive. If this happens, we
+ * nudge the WALWriter, which will take care of notifying segments as
+ * needed.
+ */
+ if (StartSeg != EndSeg && XLogArchivingActive() &&
+ LogwrtResult.Flush >= EndPos && ProcGlobal->walwriterLatch)
+ SetLatch(ProcGlobal->walwriterLatch);
}
/*
* If this was an XLOG_SWITCH record, flush the record and the empty
- * padding space that fills the rest of the segment, and perform
- * end-of-segment actions (eg, notifying archiver).
+ * padding space that fills the rest of the segment.
*/
if (isLogSwitch)
{
@@ -2421,6 +2469,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
/* We should always be inside a critical section here */
Assert(CritSectionCount > 0);
+ Assert(LWLockHeldByMe(WALWriteLock));
/*
* Update local LogwrtResult (caller probably did this already, but...)
@@ -2586,11 +2635,12 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
* later. Doing it here ensures that one and only one backend will
* perform this fsync.
*
- * This is also the right place to notify the Archiver that the
- * segment is ready to copy to archival storage, and to update the
- * timer for archive_timeout, and to signal for a checkpoint if
- * too many logfile segments have been used since the last
- * checkpoint.
+ * If WAL archiving is active, we attempt to notify the archiver
+ * of any segments that are now ready for archival.
+ *
+ * This is also the right place to update the timer for
+ * archive_timeout and to signal for a checkpoint if too many
+ * logfile segments have been used since the last checkpoint.
*/
if (finishing_seg)
{
@@ -2602,7 +2652,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
LogwrtResult.Flush = LogwrtResult.Write; /* end of page */
if (XLogArchivingActive())
- XLogArchiveNotifySeg(openLogSegNo);
+ NotifySegmentsReadyForArchive(LogwrtResult.Flush);
XLogCtl->lastSegSwitchTime = (pg_time_t) time(NULL);
XLogCtl->lastSegSwitchLSN = LogwrtResult.Flush;
@@ -2690,6 +2740,9 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
XLogCtl->LogwrtRqst.Flush = LogwrtResult.Flush;
SpinLockRelease(&XLogCtl->info_lck);
}
+
+ if (XLogArchivingActive())
+ NotifySegmentsReadyForArchive(LogwrtResult.Flush);
}
/*
@@ -4328,6 +4381,151 @@ ValidateXLOGDirectoryStructure(void)
}
}
+/*
+ * RegisterSegmentBoundary
+ * Record a new segment boundary if needed.
+ */
+static void
+RegisterSegmentBoundary(XLogSegNo seg, XLogRecPtr pos)
+{
+ XLogSegNo segno PG_USED_FOR_ASSERTS_ONLY;
+
+ /* verify caller computed segment number correctly */
+ AssertArg((XLByteToSeg(pos, segno, wal_segment_size), segno == seg));
+
+ SpinLockAcquire(&XLogCtl->segtrack_lck);
+
+ /*
+ * If no segment boundaries are registered, store the new segment boundary
+ * in earliestSegBoundary. Otherwise, store the greater segment boundaries in
+ * latestSegBoundary.
+ */
+ if (XLogCtl->earliestSegBoundary == MaxXLogSegNo)
+ {
+ XLogCtl->earliestSegBoundary = seg;
+ XLogCtl->earliestSegBoundaryRecPtr = pos;
+ }
+ else if (seg > XLogCtl->earliestSegBoundary &&
+ (XLogCtl->latestSegBoundary == MaxXLogSegNo ||
+ seg > XLogCtl->latestSegBoundary))
+ {
+ XLogCtl->latestSegBoundary = seg;
+ XLogCtl->latestSegBoundaryRecPtr = pos;
+ }
+
+ SpinLockRelease(&XLogCtl->segtrack_lck);
+}
+
+/*
+ * GetLatestSegmentBoundary
+ *
+ * Look up the latest segment boundary that is less than or equal to the
+ * given "flushed" pointer. If such a segment is found,
+ * latest_boundary_seg is populated and true is returned. Otherwise, false
+ * is returned.
+ *
+ * Callers should hold XLogCtl->segtrack_lck when calling this function.
+ */
+static bool
+GetLatestSegmentBoundary(XLogRecPtr flushed, XLogSegNo *latest_boundary_seg)
+{
+ XLogSegNo flushed_seg;
+
+ Assert(latest_boundary_seg != NULL);
+
+ XLByteToSeg(flushed, flushed_seg, wal_segment_size);
+
+ if (XLogCtl->latestSegBoundary <= flushed_seg &&
+ XLogCtl->latestSegBoundaryRecPtr <= flushed)
+ {
+ *latest_boundary_seg = XLogCtl->latestSegBoundary;
+ return true;
+ }
+
+ if (XLogCtl->earliestSegBoundary <= flushed_seg &&
+ XLogCtl->earliestSegBoundaryRecPtr <= flushed)
+ {
+ *latest_boundary_seg = XLogCtl->earliestSegBoundary;
+ return true;
+ }
+
+ return false;
+}
+
+/*
+ * RemoveSegmentBoundariesUpTo
+ *
+ * Remove all segment boundaries with segment numbers up to and including
+ * seg.
+ *
+ * Callers should hold XLogCtl->segtrack_lck when calling this function.
+ */
+static void
+RemoveSegmentBoundariesUpTo(XLogSegNo seg)
+{
+ if (XLogCtl->latestSegBoundary <= seg)
+ {
+ XLogCtl->earliestSegBoundary = MaxXLogSegNo;
+ XLogCtl->earliestSegBoundaryRecPtr = InvalidXLogRecPtr;
+
+ XLogCtl->latestSegBoundary = MaxXLogSegNo;
+ XLogCtl->latestSegBoundaryRecPtr = InvalidXLogRecPtr;
+ }
+ else if (XLogCtl->earliestSegBoundary <= seg)
+ {
+ XLogCtl->earliestSegBoundary = XLogCtl->latestSegBoundary;
+ XLogCtl->earliestSegBoundaryRecPtr = XLogCtl->latestSegBoundaryRecPtr;
+
+ XLogCtl->latestSegBoundary = MaxXLogSegNo;
+ XLogCtl->latestSegBoundaryRecPtr = InvalidXLogRecPtr;
+ }
+}
+
+/*
+ * NotifySegmentsReadyForArchive
+ *
+ * Mark segments as ready for archival, given that it is safe to do so.
+ * This function is idempotent.
+ */
+void
+NotifySegmentsReadyForArchive(XLogRecPtr flushRecPtr)
+{
+ XLogSegNo latest_boundary_seg = 0;
+
+ SpinLockAcquire(&XLogCtl->segtrack_lck);
+
+ /* Retrieve the latest segment boundary to use for notifying segments. */
+ if (GetLatestSegmentBoundary(flushRecPtr, &latest_boundary_seg))
+ {
+ XLogSegNo last_notified = XLogCtl->lastNotifiedSeg;
+
+ /*
+ * Update shared memory and discard segment boundaries that are no
+ * longer needed.
+ *
+ * It is safe to update shared memory before we attempt to create the
+ * .ready files. If our calls to XLogArchiveNotifySeg() fail,
+ * RemoveOldXlogFiles() will retry it as needed.
+ */
+ if (last_notified < latest_boundary_seg - 1)
+ XLogCtl->lastNotifiedSeg = latest_boundary_seg - 1;
+ RemoveSegmentBoundariesUpTo(latest_boundary_seg);
+
+ SpinLockRelease(&XLogCtl->segtrack_lck);
+
+ /*
+ * Notify archiver about segments that are ready for archival (by
+ * creating the corresponding .ready files).
+ */
+ for (XLogSegNo seg = last_notified + 1; seg < latest_boundary_seg; seg++)
+ XLogArchiveNotifySeg(seg, false);
+
+ PgArchWakeup();
+ }
+ else
+ SpinLockRelease(&XLogCtl->segtrack_lck);
+}
+
/*
* Remove previous backup history files. This also retries creation of
* .ready files for any backup history files for which XLogArchiveNotify
@@ -5230,9 +5428,17 @@ XLOGShmemInit(void)
SpinLockInit(&XLogCtl->Insert.insertpos_lck);
SpinLockInit(&XLogCtl->info_lck);
+ SpinLockInit(&XLogCtl->segtrack_lck);
SpinLockInit(&XLogCtl->ulsn_lck);
InitSharedLatch(&XLogCtl->recoveryWakeupLatch);
ConditionVariableInit(&XLogCtl->recoveryNotPausedCV);
+
+ /* Initialize stuff for marking segments as ready for archival. */
+ XLogCtl->lastNotifiedSeg = MaxXLogSegNo;
+ XLogCtl->earliestSegBoundary = MaxXLogSegNo;
+ XLogCtl->earliestSegBoundaryRecPtr = InvalidXLogRecPtr;
+ XLogCtl->latestSegBoundary = MaxXLogSegNo;
+ XLogCtl->latestSegBoundaryRecPtr = InvalidXLogRecPtr;
}
/*
@@ -7873,6 +8079,20 @@ StartupXLOG(void)
XLogCtl->LogwrtRqst.Write = EndOfLog;
XLogCtl->LogwrtRqst.Flush = EndOfLog;
+ /*
+ * Initialize XLogCtl->lastNotifiedSeg to the previous WAL file.
+ */
+ if (XLogArchivingActive())
+ {
+ XLogSegNo EndOfLogSeg;
+
+ XLByteToSeg(EndOfLog, EndOfLogSeg, wal_segment_size);
+
+ SpinLockAcquire(&XLogCtl->segtrack_lck);
+ XLogCtl->lastNotifiedSeg = EndOfLogSeg - 1;
+ SpinLockRelease(&XLogCtl->segtrack_lck);
+ }
+
/*
* Update full_page_writes in shared memory and write an XLOG_FPW_CHANGE
* record before resource manager writes cleanup WAL records or checkpoint
@@ -8000,7 +8220,7 @@ StartupXLOG(void)
XLogArchiveCleanup(partialfname);
durable_rename(origpath, partialpath, ERROR);
- XLogArchiveNotify(partialfname);
+ XLogArchiveNotify(partialfname, true);
}
}
}
diff --git a/src/backend/access/transam/xlogarchive.c b/src/backend/access/transam/xlogarchive.c
index 26b023e754..b9c19b2085 100644
--- a/src/backend/access/transam/xlogarchive.c
+++ b/src/backend/access/transam/xlogarchive.c
@@ -433,7 +433,7 @@ KeepFileRestoredFromArchive(const char *path, const char *xlogfname)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(xlogfname);
else
- XLogArchiveNotify(xlogfname);
+ XLogArchiveNotify(xlogfname, true);
/*
* If the existing file was replaced, since walsenders might have it open,
@@ -462,9 +462,12 @@ KeepFileRestoredFromArchive(const char *path, const char *xlogfname)
* by the archiver, e.g. we write 0000000100000001000000C6.ready
* and the archiver then knows to archive XLOGDIR/0000000100000001000000C6,
* then when complete, rename it to 0000000100000001000000C6.done
+ *
+ * Optionally, nudge the archiver process so that it'll notice the file we
+ * create.
*/
void
-XLogArchiveNotify(const char *xlog)
+XLogArchiveNotify(const char *xlog, bool nudge)
{
char archiveStatusPath[MAXPGPATH];
FILE *fd;
@@ -489,8 +492,8 @@ XLogArchiveNotify(const char *xlog)
return;
}
- /* Notify archiver that it's got something to do */
- if (IsUnderPostmaster)
+ /* If caller requested, let archiver know it's got work to do */
+ if (nudge)
PgArchWakeup();
}
@@ -498,12 +501,12 @@ XLogArchiveNotify(const char *xlog)
* Convenience routine to notify using segment number representation of filename
*/
void
-XLogArchiveNotifySeg(XLogSegNo segno)
+XLogArchiveNotifySeg(XLogSegNo segno, bool nudge)
{
char xlog[MAXFNAMELEN];
XLogFileName(xlog, ThisTimeLineID, segno, wal_segment_size);
- XLogArchiveNotify(xlog);
+ XLogArchiveNotify(xlog, nudge);
}
/*
@@ -608,7 +611,7 @@ XLogArchiveCheckDone(const char *xlog)
return true;
/* Retry creation of the .ready file */
- XLogArchiveNotify(xlog);
+ XLogArchiveNotify(xlog, true);
return false;
}
diff --git a/src/backend/postmaster/walwriter.c b/src/backend/postmaster/walwriter.c
index 626fae8454..6a1e16edc2 100644
--- a/src/backend/postmaster/walwriter.c
+++ b/src/backend/postmaster/walwriter.c
@@ -248,6 +248,13 @@ WalWriterMain(void)
/* Process any signals received recently */
HandleWalWriterInterrupts();
+ /*
+ * Notify the archiver of any WAL segments that are ready. We do this
+ * here to handle a race condition where WAL is flushed to disk prior
+ * to registering the segment boundary.
+ */
+ NotifySegmentsReadyForArchive(GetFlushRecPtr());
+
/*
* Do what we're here for; then, if XLogBackgroundFlush() found useful
* work to do, reset hibernation counter.
diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c
index 9a2bc37fd7..60de3be92c 100644
--- a/src/backend/replication/walreceiver.c
+++ b/src/backend/replication/walreceiver.c
@@ -622,7 +622,7 @@ WalReceiverMain(void)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(xlogfname);
else
- XLogArchiveNotify(xlogfname);
+ XLogArchiveNotify(xlogfname, true);
}
recvFile = -1;
@@ -760,7 +760,7 @@ WalRcvFetchTimeLineHistoryFiles(TimeLineID first, TimeLineID last)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(fname);
else
- XLogArchiveNotify(fname);
+ XLogArchiveNotify(fname, true);
pfree(fname);
pfree(content);
@@ -915,7 +915,7 @@ XLogWalRcvWrite(char *buf, Size nbytes, XLogRecPtr recptr)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(xlogfname);
else
- XLogArchiveNotify(xlogfname);
+ XLogArchiveNotify(xlogfname, true);
}
recvFile = -1;
diff --git a/src/include/access/xlog.h b/src/include/access/xlog.h
index 0a8ede700d..6b6ae81c2d 100644
--- a/src/include/access/xlog.h
+++ b/src/include/access/xlog.h
@@ -315,6 +315,7 @@ extern XLogRecPtr GetInsertRecPtr(void);
extern XLogRecPtr GetFlushRecPtr(void);
extern XLogRecPtr GetLastImportantRecPtr(void);
extern void RemovePromoteSignalFiles(void);
+extern void NotifySegmentsReadyForArchive(XLogRecPtr flushRecPtr);
extern bool PromoteIsTriggered(void);
extern bool CheckPromoteSignal(void);
diff --git a/src/include/access/xlogarchive.h b/src/include/access/xlogarchive.h
index 3edd1a976c..935b4cb02d 100644
--- a/src/include/access/xlogarchive.h
+++ b/src/include/access/xlogarchive.h
@@ -23,8 +23,8 @@ extern bool RestoreArchivedFile(char *path, const char *xlogfname,
extern void ExecuteRecoveryCommand(const char *command, const char *commandName,
bool failOnSignal);
extern void KeepFileRestoredFromArchive(const char *path, const char *xlogfname);
-extern void XLogArchiveNotify(const char *xlog);
-extern void XLogArchiveNotifySeg(XLogSegNo segno);
+extern void XLogArchiveNotify(const char *xlog, bool nudge);
+extern void XLogArchiveNotifySeg(XLogSegNo segno, bool nudge);
extern void XLogArchiveForceDone(const char *xlog);
extern bool XLogArchiveCheckDone(const char *xlog);
extern bool XLogArchiveIsBusy(const char *xlog);
diff --git a/src/include/access/xlogdefs.h b/src/include/access/xlogdefs.h
index 60348d1850..9b455e88e3 100644
--- a/src/include/access/xlogdefs.h
+++ b/src/include/access/xlogdefs.h
@@ -46,6 +46,7 @@ typedef uint64 XLogRecPtr;
* XLogSegNo - physical log file sequence number.
*/
typedef uint64 XLogSegNo;
+#define MaxXLogSegNo ((XLogSegNo) 0xFFFFFFFFFFFFFFFF)
/*
* TimeLineID (TLI) - identifies different database histories to prevent
--
2.30.2
--3l72er6sf3zeysw4--
^ permalink raw reply [nested|flat] 19+ messages in thread
* [PATCH v15] Avoid creating archive status ".ready" files too early
@ 2021-08-23 13:06 Alvaro Herrera <[email protected]>
0 siblings, 0 replies; 19+ messages in thread
From: Alvaro Herrera @ 2021-08-23 13:06 UTC (permalink / raw)
WAL records may span multiple segments, but XLogWrite() does not
wait for the entire record to be written out to disk before
creating archive status files. Instead, as soon as the last WAL page of
the segment is written, the archive status file is created, and the
archiver may process it. If PostgreSQL crashes before it is able to
write and flush the rest of the record (in the next WAL segment), the
wrong version of the first segment file lingers in the archive, which
causes operations such as point-in-time restores to fail.
To fix this, keep track of records that span across segments and ensure
that segments are only marked ready-for-archival once such records have
been completely written to disk.
This has always been wrong, so backpatch all the way back.
Author: Nathan Bossart <[email protected]>
Reviewed-by: Kyotaro Horiguchi <[email protected]>
Reviewed-by: Ryo Matsumura <[email protected]>
Reviewed-by: Andrey Borodin <[email protected]>
Discussion: https://postgr.es/m/[email protected]
---
src/backend/access/transam/timeline.c | 2 +-
src/backend/access/transam/xlog.c | 220 +++++++++++++++++++++--
src/backend/access/transam/xlogarchive.c | 17 +-
src/backend/postmaster/walwriter.c | 7 +
src/backend/replication/walreceiver.c | 6 +-
src/include/access/xlog.h | 1 +
src/include/access/xlogarchive.h | 4 +-
src/include/access/xlogdefs.h | 1 +
8 files changed, 234 insertions(+), 24 deletions(-)
diff --git a/src/backend/access/transam/timeline.c b/src/backend/access/transam/timeline.c
index 8d0903c175..acd5c2431d 100644
--- a/src/backend/access/transam/timeline.c
+++ b/src/backend/access/transam/timeline.c
@@ -452,7 +452,7 @@ writeTimeLineHistory(TimeLineID newTLI, TimeLineID parentTLI,
if (XLogArchivingActive())
{
TLHistoryFileName(histfname, newTLI);
- XLogArchiveNotify(histfname);
+ XLogArchiveNotify(histfname, true);
}
}
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e51a7a749d..8e7c3a364a 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -724,6 +724,18 @@ typedef struct XLogCtlData
XLogRecPtr lastFpwDisableRecPtr;
slock_t info_lck; /* locks shared variables shown above */
+
+ /*
+ * Variables used to track cross-segment records. Protected by
+ * segtrack_lck.
+ */
+ XLogSegNo lastNotifiedSeg;
+ XLogSegNo earliestSegBoundary;
+ XLogRecPtr earliestSegBoundaryRecPtr;
+ XLogSegNo latestSegBoundary;
+ XLogRecPtr latestSegBoundaryRecPtr;
+
+ slock_t segtrack_lck; /* locks shared variables shown above */
} XLogCtlData;
static XLogCtlData *XLogCtl = NULL;
@@ -920,6 +932,7 @@ static void RemoveXlogFile(const char *segname, XLogSegNo recycleSegNo,
XLogSegNo *endlogSegNo);
static void UpdateLastRemovedPtr(char *filename);
static void ValidateXLOGDirectoryStructure(void);
+static void RegisterSegmentBoundary(XLogSegNo seg, XLogRecPtr pos);
static void CleanupBackupHistory(void);
static void UpdateMinRecoveryPoint(XLogRecPtr lsn, bool force);
static XLogRecord *ReadRecord(XLogReaderState *xlogreader,
@@ -1154,23 +1167,56 @@ XLogInsertRecord(XLogRecData *rdata,
END_CRIT_SECTION();
/*
- * Update shared LogwrtRqst.Write, if we crossed page boundary.
+ * If we crossed page boundary, update LogwrtRqst.Write; if we crossed
+ * segment boundary, register that and wake up walwriter.
*/
if (StartPos / XLOG_BLCKSZ != EndPos / XLOG_BLCKSZ)
{
+ XLogSegNo StartSeg;
+ XLogSegNo EndSeg;
+
+ XLByteToSeg(StartPos, StartSeg, wal_segment_size);
+ XLByteToSeg(EndPos, EndSeg, wal_segment_size);
+
+ /*
+ * Register our crossing the segment boundary if that occurred.
+ *
+ * Note that we did not use XLByteToPrevSeg() for determining the
+ * ending segment. This is so that a record that fits perfectly into
+ * the end of the segment is marked ready for archival as soon as the
+ * flushed pointer jumps to the next segment.
+ */
+ if (StartSeg != EndSeg && XLogArchivingActive())
+ RegisterSegmentBoundary(EndSeg, EndPos);
+
+ /*
+ * Advance LogwrtRqst.Write so that it includes new block(s).
+ *
+ * We do this after registering the segment boundary so that the
+ * comparison with the flushed pointer below can use the latest value
+ * known globally.
+ */
SpinLockAcquire(&XLogCtl->info_lck);
- /* advance global request to include new block(s) */
if (XLogCtl->LogwrtRqst.Write < EndPos)
XLogCtl->LogwrtRqst.Write = EndPos;
/* update local result copy while I have the chance */
LogwrtResult = XLogCtl->LogwrtResult;
SpinLockRelease(&XLogCtl->info_lck);
+
+ /*
+ * There's a chance that the record was already flushed to disk and we
+ * missed marking segments as ready for archive. If this happens, we
+ * nudge the WALWriter, which will take care of notifying segments as
+ * needed.
+ */
+ if (StartSeg != EndSeg && XLogArchivingActive() &&
+ LogwrtResult.Flush >= EndPos && ProcGlobal->walwriterLatch)
+ SetLatch(ProcGlobal->walwriterLatch);
}
/*
* If this was an XLOG_SWITCH record, flush the record and the empty
- * padding space that fills the rest of the segment, and perform
- * end-of-segment actions (eg, notifying archiver).
+ * padding space that fills the rest of the segment.
*/
if (isLogSwitch)
{
@@ -2421,6 +2467,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
/* We should always be inside a critical section here */
Assert(CritSectionCount > 0);
+ Assert(LWLockHeldByMe(WALWriteLock));
/*
* Update local LogwrtResult (caller probably did this already, but...)
@@ -2586,11 +2633,12 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
* later. Doing it here ensures that one and only one backend will
* perform this fsync.
*
- * This is also the right place to notify the Archiver that the
- * segment is ready to copy to archival storage, and to update the
- * timer for archive_timeout, and to signal for a checkpoint if
- * too many logfile segments have been used since the last
- * checkpoint.
+ * If WAL archiving is active, we attempt to notify the archiver
+ * of any segments that are now ready for archival.
+ *
+ * This is also the right place to update the timer for
+ * archive_timeout and to signal for a checkpoint if too many
+ * logfile segments have been used since the last checkpoint.
*/
if (finishing_seg)
{
@@ -2602,7 +2650,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
LogwrtResult.Flush = LogwrtResult.Write; /* end of page */
if (XLogArchivingActive())
- XLogArchiveNotifySeg(openLogSegNo);
+ NotifySegmentsReadyForArchive(LogwrtResult.Flush);
XLogCtl->lastSegSwitchTime = (pg_time_t) time(NULL);
XLogCtl->lastSegSwitchLSN = LogwrtResult.Flush;
@@ -2690,6 +2738,9 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
XLogCtl->LogwrtRqst.Flush = LogwrtResult.Flush;
SpinLockRelease(&XLogCtl->info_lck);
}
+
+ if (XLogArchivingActive())
+ NotifySegmentsReadyForArchive(LogwrtResult.Flush);
}
/*
@@ -4328,6 +4379,131 @@ ValidateXLOGDirectoryStructure(void)
}
}
+/*
+ * RegisterSegmentBoundary
+ *
+ * WAL records that are split across a segment boundary require special
+ * treatment for archiving: the initial segment must not be archived until
+ * the end segment has been flushed, in case we crash before we have
+ * the chance to flush the end segment (because after recovery we would
+ * overwrite that WAL record with a different one, and so the file we
+ * archived no longer represents truth.) This also applies to streaming
+ * physical replication.
+ *
+ * To handle this, we keep track of the LSN of WAL records that cross
+ * segment boundaries. Two such are sufficient: the earliest and the
+ * latest we know about, since the flush position advances monotonically.
+ * WAL record writers register boundary-crossing records here, which is
+ * used by .ready file creation to delay until the end segment is known
+ * flushed.
+ */
+static void
+RegisterSegmentBoundary(XLogSegNo seg, XLogRecPtr pos)
+{
+ XLogSegNo segno PG_USED_FOR_ASSERTS_ONLY;
+
+ /* verify caller computed segment number correctly */
+ AssertArg((XLByteToSeg(pos, segno, wal_segment_size), segno == seg));
+
+ SpinLockAcquire(&XLogCtl->segtrack_lck);
+
+ /*
+ * If no segment boundaries are registered, store the new segment boundary
+ * in earliestSegBoundary. Otherwise, store the greater segment boundaries in
+ * latestSegBoundary.
+ */
+ if (XLogCtl->earliestSegBoundary == MaxXLogSegNo)
+ {
+ XLogCtl->earliestSegBoundary = seg;
+ XLogCtl->earliestSegBoundaryRecPtr = pos;
+ }
+ else if (seg > XLogCtl->earliestSegBoundary &&
+ (XLogCtl->latestSegBoundary == MaxXLogSegNo ||
+ seg > XLogCtl->latestSegBoundary))
+ {
+ XLogCtl->latestSegBoundary = seg;
+ XLogCtl->latestSegBoundaryRecPtr = pos;
+ }
+
+ SpinLockRelease(&XLogCtl->segtrack_lck);
+}
+
+/*
+ * NotifySegmentsReadyForArchive
+ *
+ * Mark segments as ready for archival, given that it is safe to do so.
+ * This function is idempotent.
+ */
+void
+NotifySegmentsReadyForArchive(XLogRecPtr flushRecPtr)
+{
+ XLogSegNo latest_boundary_seg;
+ XLogSegNo last_notified;
+ XLogSegNo flushed_seg;
+ XLogSegNo seg;
+ bool keep_latest;
+
+ XLByteToSeg(flushRecPtr, flushed_seg, wal_segment_size);
+
+ SpinLockAcquire(&XLogCtl->segtrack_lck);
+
+ if (XLogCtl->latestSegBoundary <= flushed_seg &&
+ XLogCtl->latestSegBoundaryRecPtr <= flushRecPtr)
+ {
+ latest_boundary_seg = XLogCtl->latestSegBoundary;
+ keep_latest = false;
+ }
+ else if (XLogCtl->earliestSegBoundary <= flushed_seg &&
+ XLogCtl->earliestSegBoundaryRecPtr <= flushRecPtr)
+ {
+ latest_boundary_seg = XLogCtl->earliestSegBoundary;
+ keep_latest = true;
+ }
+ else
+ {
+ SpinLockRelease(&XLogCtl->segtrack_lck);
+ return;
+ }
+
+ last_notified = XLogCtl->lastNotifiedSeg;
+
+ /*
+ * Update shared memory and discard segment boundaries that are no
+ * longer needed.
+ *
+ * It is safe to update shared memory before we attempt to create the
+ * .ready files. If our calls to XLogArchiveNotifySeg() fail,
+ * RemoveOldXlogFiles() will retry it as needed.
+ */
+ if (last_notified < latest_boundary_seg - 1)
+ XLogCtl->lastNotifiedSeg = latest_boundary_seg - 1;
+
+ if (keep_latest)
+ {
+ XLogCtl->earliestSegBoundary = XLogCtl->latestSegBoundary;
+ XLogCtl->earliestSegBoundaryRecPtr = XLogCtl->latestSegBoundaryRecPtr;
+ }
+ else
+ {
+ XLogCtl->earliestSegBoundary = MaxXLogSegNo;
+ XLogCtl->earliestSegBoundaryRecPtr = InvalidXLogRecPtr;
+ }
+
+ XLogCtl->latestSegBoundary = MaxXLogSegNo;
+ XLogCtl->latestSegBoundaryRecPtr = InvalidXLogRecPtr;
+
+ SpinLockRelease(&XLogCtl->segtrack_lck);
+
+ /*
+ * Notify archiver about segments that are ready for archival (by
+ * creating the corresponding .ready files).
+ */
+ for (seg = last_notified + 1; seg < latest_boundary_seg; seg++)
+ XLogArchiveNotifySeg(seg, false);
+
+ PgArchWakeup();
+}
+
/*
* Remove previous backup history files. This also retries creation of
* .ready files for any backup history files for which XLogArchiveNotify
@@ -5230,9 +5406,17 @@ XLOGShmemInit(void)
SpinLockInit(&XLogCtl->Insert.insertpos_lck);
SpinLockInit(&XLogCtl->info_lck);
+ SpinLockInit(&XLogCtl->segtrack_lck);
SpinLockInit(&XLogCtl->ulsn_lck);
InitSharedLatch(&XLogCtl->recoveryWakeupLatch);
ConditionVariableInit(&XLogCtl->recoveryNotPausedCV);
+
+ /* Initialize stuff for marking segments as ready for archival. */
+ XLogCtl->lastNotifiedSeg = MaxXLogSegNo;
+ XLogCtl->earliestSegBoundary = MaxXLogSegNo;
+ XLogCtl->earliestSegBoundaryRecPtr = InvalidXLogRecPtr;
+ XLogCtl->latestSegBoundary = MaxXLogSegNo;
+ XLogCtl->latestSegBoundaryRecPtr = InvalidXLogRecPtr;
}
/*
@@ -7873,6 +8057,20 @@ StartupXLOG(void)
XLogCtl->LogwrtRqst.Write = EndOfLog;
XLogCtl->LogwrtRqst.Flush = EndOfLog;
+ /*
+ * Initialize XLogCtl->lastNotifiedSeg to the previous WAL file.
+ */
+ if (XLogArchivingActive())
+ {
+ XLogSegNo EndOfLogSeg;
+
+ XLByteToSeg(EndOfLog, EndOfLogSeg, wal_segment_size);
+
+ SpinLockAcquire(&XLogCtl->segtrack_lck);
+ XLogCtl->lastNotifiedSeg = EndOfLogSeg - 1;
+ SpinLockRelease(&XLogCtl->segtrack_lck);
+ }
+
/*
* Update full_page_writes in shared memory and write an XLOG_FPW_CHANGE
* record before resource manager writes cleanup WAL records or checkpoint
@@ -8000,7 +8198,7 @@ StartupXLOG(void)
XLogArchiveCleanup(partialfname);
durable_rename(origpath, partialpath, ERROR);
- XLogArchiveNotify(partialfname);
+ XLogArchiveNotify(partialfname, true);
}
}
}
diff --git a/src/backend/access/transam/xlogarchive.c b/src/backend/access/transam/xlogarchive.c
index 26b023e754..b9c19b2085 100644
--- a/src/backend/access/transam/xlogarchive.c
+++ b/src/backend/access/transam/xlogarchive.c
@@ -433,7 +433,7 @@ KeepFileRestoredFromArchive(const char *path, const char *xlogfname)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(xlogfname);
else
- XLogArchiveNotify(xlogfname);
+ XLogArchiveNotify(xlogfname, true);
/*
* If the existing file was replaced, since walsenders might have it open,
@@ -462,9 +462,12 @@ KeepFileRestoredFromArchive(const char *path, const char *xlogfname)
* by the archiver, e.g. we write 0000000100000001000000C6.ready
* and the archiver then knows to archive XLOGDIR/0000000100000001000000C6,
* then when complete, rename it to 0000000100000001000000C6.done
+ *
+ * Optionally, nudge the archiver process so that it'll notice the file we
+ * create.
*/
void
-XLogArchiveNotify(const char *xlog)
+XLogArchiveNotify(const char *xlog, bool nudge)
{
char archiveStatusPath[MAXPGPATH];
FILE *fd;
@@ -489,8 +492,8 @@ XLogArchiveNotify(const char *xlog)
return;
}
- /* Notify archiver that it's got something to do */
- if (IsUnderPostmaster)
+ /* If caller requested, let archiver know it's got work to do */
+ if (nudge)
PgArchWakeup();
}
@@ -498,12 +501,12 @@ XLogArchiveNotify(const char *xlog)
* Convenience routine to notify using segment number representation of filename
*/
void
-XLogArchiveNotifySeg(XLogSegNo segno)
+XLogArchiveNotifySeg(XLogSegNo segno, bool nudge)
{
char xlog[MAXFNAMELEN];
XLogFileName(xlog, ThisTimeLineID, segno, wal_segment_size);
- XLogArchiveNotify(xlog);
+ XLogArchiveNotify(xlog, nudge);
}
/*
@@ -608,7 +611,7 @@ XLogArchiveCheckDone(const char *xlog)
return true;
/* Retry creation of the .ready file */
- XLogArchiveNotify(xlog);
+ XLogArchiveNotify(xlog, true);
return false;
}
diff --git a/src/backend/postmaster/walwriter.c b/src/backend/postmaster/walwriter.c
index 626fae8454..6a1e16edc2 100644
--- a/src/backend/postmaster/walwriter.c
+++ b/src/backend/postmaster/walwriter.c
@@ -248,6 +248,13 @@ WalWriterMain(void)
/* Process any signals received recently */
HandleWalWriterInterrupts();
+ /*
+ * Notify the archiver of any WAL segments that are ready. We do this
+ * here to handle a race condition where WAL is flushed to disk prior
+ * to registering the segment boundary.
+ */
+ NotifySegmentsReadyForArchive(GetFlushRecPtr());
+
/*
* Do what we're here for; then, if XLogBackgroundFlush() found useful
* work to do, reset hibernation counter.
diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c
index 9a2bc37fd7..60de3be92c 100644
--- a/src/backend/replication/walreceiver.c
+++ b/src/backend/replication/walreceiver.c
@@ -622,7 +622,7 @@ WalReceiverMain(void)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(xlogfname);
else
- XLogArchiveNotify(xlogfname);
+ XLogArchiveNotify(xlogfname, true);
}
recvFile = -1;
@@ -760,7 +760,7 @@ WalRcvFetchTimeLineHistoryFiles(TimeLineID first, TimeLineID last)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(fname);
else
- XLogArchiveNotify(fname);
+ XLogArchiveNotify(fname, true);
pfree(fname);
pfree(content);
@@ -915,7 +915,7 @@ XLogWalRcvWrite(char *buf, Size nbytes, XLogRecPtr recptr)
if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
XLogArchiveForceDone(xlogfname);
else
- XLogArchiveNotify(xlogfname);
+ XLogArchiveNotify(xlogfname, true);
}
recvFile = -1;
diff --git a/src/include/access/xlog.h b/src/include/access/xlog.h
index 0a8ede700d..6b6ae81c2d 100644
--- a/src/include/access/xlog.h
+++ b/src/include/access/xlog.h
@@ -315,6 +315,7 @@ extern XLogRecPtr GetInsertRecPtr(void);
extern XLogRecPtr GetFlushRecPtr(void);
extern XLogRecPtr GetLastImportantRecPtr(void);
extern void RemovePromoteSignalFiles(void);
+extern void NotifySegmentsReadyForArchive(XLogRecPtr flushRecPtr);
extern bool PromoteIsTriggered(void);
extern bool CheckPromoteSignal(void);
diff --git a/src/include/access/xlogarchive.h b/src/include/access/xlogarchive.h
index 3edd1a976c..935b4cb02d 100644
--- a/src/include/access/xlogarchive.h
+++ b/src/include/access/xlogarchive.h
@@ -23,8 +23,8 @@ extern bool RestoreArchivedFile(char *path, const char *xlogfname,
extern void ExecuteRecoveryCommand(const char *command, const char *commandName,
bool failOnSignal);
extern void KeepFileRestoredFromArchive(const char *path, const char *xlogfname);
-extern void XLogArchiveNotify(const char *xlog);
-extern void XLogArchiveNotifySeg(XLogSegNo segno);
+extern void XLogArchiveNotify(const char *xlog, bool nudge);
+extern void XLogArchiveNotifySeg(XLogSegNo segno, bool nudge);
extern void XLogArchiveForceDone(const char *xlog);
extern bool XLogArchiveCheckDone(const char *xlog);
extern bool XLogArchiveIsBusy(const char *xlog);
diff --git a/src/include/access/xlogdefs.h b/src/include/access/xlogdefs.h
index 60348d1850..9b455e88e3 100644
--- a/src/include/access/xlogdefs.h
+++ b/src/include/access/xlogdefs.h
@@ -46,6 +46,7 @@ typedef uint64 XLogRecPtr;
* XLogSegNo - physical log file sequence number.
*/
typedef uint64 XLogSegNo;
+#define MaxXLogSegNo ((XLogSegNo) 0xFFFFFFFFFFFFFFFF)
/*
* TimeLineID (TLI) - identifies different database histories to prevent
--
2.30.2
--erxcxkqg4k65gkj3--
^ permalink raw reply [nested|flat] 19+ messages in thread
* Re: Option to not use ringbuffer in VACUUM, using it in failsafe mode
@ 2023-01-12 00:26 David G. Johnston <[email protected]>
2023-01-12 00:36 ` Re: Option to not use ringbuffer in VACUUM, using it in failsafe mode Andres Freund <[email protected]>
0 siblings, 1 reply; 19+ messages in thread
From: David G. Johnston @ 2023-01-12 00:26 UTC (permalink / raw)
To: Andres Freund <[email protected]>; +Cc: Tom Lane <[email protected]>; Peter Geoghegan <[email protected]>; pgsql-hackers; Robert Haas <[email protected]>; Melanie Plageman <[email protected]>; David Rowley <[email protected]>
On Wed, Jan 11, 2023 at 2:39 PM Andres Freund <[email protected]> wrote:
> Hi,
>
> On 2023-01-11 16:18:34 -0500, Tom Lane wrote:
> > Peter Geoghegan <[email protected]> writes:
> > > On Wed, Jan 11, 2023 at 11:18 AM Andres Freund <[email protected]>
> wrote:
> > >> I don't like that - it's also quite useful to disable use of
> ringbuffers when
> > >> you actually need to clean up indexes. Especially when we have a lot
> of dead
> > >> tuples we'll rescan indexes over and over...
> >
> > > That's a fair point.
> >
> > > My vote goes to "REUSE_BUFFERS", then.
> >
> > I wonder whether it could make sense to allow a larger ringbuffer size,
> > rather than just the limit cases of "on" and "off".
>
> I can see that making sense, particularly if we were to later extend this
> to
> other users of ringbuffers. E.g. COPYs us of the ringbuffer makes loading
> of
> data > 16MB but also << s_b vastly slower, but it can still be very
> important
> to use if there's lots of parallel processes loading data.
>
>
Should we just add "ring_buffers" to the existing "shared_buffers" and
"temp_buffers" settings?
Then give VACUUM a (BUFFER_POOL=ring*|shared) option?
I think making DBAs aware of this dynamic and making the ring buffer usage
user-facing is beneficial in its own right (at least, the concept that
changes done by vacuum don't impact shared_buffers, regardless of how that
non-impact manifests). But I don't see much benefit trying to come up with
a different name.
David J.
^ permalink raw reply [nested|flat] 19+ messages in thread
* Re: Option to not use ringbuffer in VACUUM, using it in failsafe mode
2023-01-12 00:26 Re: Option to not use ringbuffer in VACUUM, using it in failsafe mode David G. Johnston <[email protected]>
@ 2023-01-12 00:36 ` Andres Freund <[email protected]>
2023-02-28 08:16 ` Re: Option to not use ringbuffer in VACUUM, using it in failsafe mode Bharath Rupireddy <[email protected]>
0 siblings, 1 reply; 19+ messages in thread
From: Andres Freund @ 2023-01-12 00:36 UTC (permalink / raw)
To: David G. Johnston <[email protected]>; +Cc: Tom Lane <[email protected]>; Peter Geoghegan <[email protected]>; pgsql-hackers; Robert Haas <[email protected]>; Melanie Plageman <[email protected]>; David Rowley <[email protected]>
Hi,
On 2023-01-11 17:26:19 -0700, David G. Johnston wrote:
> Should we just add "ring_buffers" to the existing "shared_buffers" and
> "temp_buffers" settings?
The different types of ring buffers have different sizes, for good reasons. So
I don't see that working well. I also think it'd be more often useful to
control this on a statement basis - if you have a parallel import tool that
starts NCPU COPYs you'd want a smaller buffer than a single threaded COPY. Of
course each session can change the ring buffer settings, but still.
> Then give VACUUM a (BUFFER_POOL=ring*|shared) option?
That seems likely to mislead, because it'd still use shared buffers when the
blocks are already present. The ring buffers aren't a separate buffer pool,
they're a subset of the normal bufferpool. Lookup is done normally, only when
a page isn't found, the search for a victim buffer first tries to use a buffer
from the ring.
> I think making DBAs aware of this dynamic and making the ring buffer usage
> user-facing is beneficial in its own right (at least, the concept that
> changes done by vacuum don't impact shared_buffers, regardless of how that
> non-impact manifests).
VACUUM can end up dirtying all of shared buffers, even with the ring buffer in
use...
Greetings,
Andres Freund
^ permalink raw reply [nested|flat] 19+ messages in thread
* Re: Option to not use ringbuffer in VACUUM, using it in failsafe mode
2023-01-12 00:26 Re: Option to not use ringbuffer in VACUUM, using it in failsafe mode David G. Johnston <[email protected]>
2023-01-12 00:36 ` Re: Option to not use ringbuffer in VACUUM, using it in failsafe mode Andres Freund <[email protected]>
@ 2023-02-28 08:16 ` Bharath Rupireddy <[email protected]>
0 siblings, 0 replies; 19+ messages in thread
From: Bharath Rupireddy @ 2023-02-28 08:16 UTC (permalink / raw)
To: Andres Freund <[email protected]>; +Cc: David G. Johnston <[email protected]>; Tom Lane <[email protected]>; Peter Geoghegan <[email protected]>; pgsql-hackers; Robert Haas <[email protected]>; Melanie Plageman <[email protected]>; David Rowley <[email protected]>
On Thu, Jan 12, 2023 at 6:06 AM Andres Freund <[email protected]> wrote:
>
> On 2023-01-11 17:26:19 -0700, David G. Johnston wrote:
> > Should we just add "ring_buffers" to the existing "shared_buffers" and
> > "temp_buffers" settings?
>
> The different types of ring buffers have different sizes, for good reasons. So
> I don't see that working well. I also think it'd be more often useful to
> control this on a statement basis - if you have a parallel import tool that
> starts NCPU COPYs you'd want a smaller buffer than a single threaded COPY. Of
> course each session can change the ring buffer settings, but still.
How about having GUCs for each ring buffer (bulk_read_ring_buffers,
bulk_write_ring_buffers, vacuum_ring_buffers - ah, 3 more new GUCs)?
These options can help especially when statement level controls aren't
easy to add (COPY, CREATE TABLE AS/CTAS, REFRESH MAT VIEW/RMV)? If
needed users can also set them at the system level. For instance, one
can set bulk_write_ring_buffers to other than 16MB or -1 to disable
the ring buffer to use shared_buffers and run a bunch of bulk write
queries.
Although I'm not quite opposing the idea of statement level controls
(like the VACUUM one proposed here), it is better to make these ring
buffer sizes configurable across the system to help with the other
similar cases e.g., a CTAS or RMV can help subsequent reads from
shared buffers if ring buffer is skipped.
Thoughts?
--
Bharath Rupireddy
PostgreSQL Contributors Team
RDS Open Source Databases
Amazon Web Services: https://aws.amazon.com
^ permalink raw reply [nested|flat] 19+ messages in thread
end of thread, other threads:[~2023-02-28 08:16 UTC | newest]
Thread overview: 19+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2017-11-14 15:33 Re: [HACKERS] log_destination=file Robert Haas <[email protected]>
2017-11-30 04:44 ` Michael Paquier <[email protected]>
2018-01-20 12:51 ` Magnus Hagander <[email protected]>
2018-01-22 19:52 ` Robert Haas <[email protected]>
2018-01-22 21:25 ` Tomas Vondra <[email protected]>
2018-03-02 18:22 ` Robert Haas <[email protected]>
2018-03-02 18:29 ` Tom Lane <[email protected]>
2018-03-02 23:06 ` Magnus Hagander <[email protected]>
2018-03-02 07:27 ` Andres Freund <[email protected]>
2018-03-02 08:42 ` Magnus Hagander <[email protected]>
2021-07-30 22:35 [PATCH v2] Avoid creating archive status ".ready" files too early. Alvaro Herrera <[email protected]>
2021-07-30 22:35 [PATCH v3] Avoid creating archive status ".ready" files too early. Alvaro Herrera <[email protected]>
2021-08-17 03:52 [PATCH v10] Avoid creating archive status ".ready" files too early. Nathan Bossart <[email protected]>
2021-08-17 03:52 [PATCH v12] Avoid creating archive status ".ready" files too early Nathan Bossart <[email protected]>
2021-08-20 19:25 [PATCH v14] Avoid creating archive status ".ready" files too early Nathan Bossart <[email protected]>
2021-08-23 13:06 [PATCH v15] Avoid creating archive status ".ready" files too early Alvaro Herrera <[email protected]>
2023-01-12 00:26 Re: Option to not use ringbuffer in VACUUM, using it in failsafe mode David G. Johnston <[email protected]>
2023-01-12 00:36 ` Re: Option to not use ringbuffer in VACUUM, using it in failsafe mode Andres Freund <[email protected]>
2023-02-28 08:16 ` Re: Option to not use ringbuffer in VACUUM, using it in failsafe mode Bharath Rupireddy <[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