Обсуждение: pg_wal_summary_contents() and pg_walsummary may return different results on the same WAL summary file
Hi,
I found that pg_wal_summary_contents() may miss some results that pg_walsummary returns for the same WAL summary file.
Hereare the steps to reproduce the issue:
-----------------------------
initdb -D data
echo "summarize_wal = on" >> data/postgresql.conf
pg_ctl -D data start
psql <<EOF
CREATE TABLE t AS SELECT n i, n j FROM generate_series(1, 1000) n;
DELETE FROM t;
CHECKPOINT;
VACUUM t;
CHECKPOINT;
SELECT foo.* FROM (SELECT * FROM pg_available_wal_summaries() ORDER BY start_lsn DESC LIMIT 1) JOIN LATERAL
pg_wal_summary_contents(tli,start_lsn, end_lsn) foo ON true;
EOF
pg_walsummary -i data/pg_wal/summaries/$(ls -1 data/pg_wal/summaries/ | tail -1)
-----------------------------
In my test, pg_walsummary returned three records:
TS 1663, DB 5, REL 1259, FORK main: block 0
TS 1663, DB 5, REL 16384, FORK main: limit 0
TS 1663, DB 5, REL 16384, FORK vm: limit 0
However, pg_wal_summary_contents() returned only one record:
relfilenode | reltablespace | reldatabase | relforknumber | relblocknumber | is_limit_block
-------------+---------------+-------------+---------------+----------------+----------------
1259 | 1663 | 5 | 0 | 0 | f
pg_wal_summary_contents() seems to miss the summary information with "limit" that pg_walsummary reports. This appears
tobe a bug. The attached patch fixes this.
By the way, pg_wal_summary_contents() and pg_walsummary perform nearly the same task but are implemented in different
functions.This could be the root of issues like this. In the future, it would be better to have a common function for
outputtingthe WAL summary file that both can use.
Regards,
--
Fujii Masao
Advanced Computing Technology Center
Research and Development Headquarters
NTT DATA CORPORATION
Вложения
On Wed, Jul 3, 2024 at 5:34 AM Fujii Masao <masao.fujii@oss.nttdata.com> wrote: > pg_wal_summary_contents() seems to miss the summary information with "limit" that pg_walsummary reports. This appears tobe a bug. The attached patch fixes this. Oops. It looks like pg_wal_summary_contents() forgets to emit the limit block when that's the only data for a particular relation fork. And maybe you can make it emit the limit block multiple times if the list of block numbers is long enough. Thanks for the patch. I think you can commit and back-patch this, but I don't think the commit message is quite right, because it's not like this code just NEVER executes where it is located currently. Or am I missing something? > By the way, pg_wal_summary_contents() and pg_walsummary perform nearly the same task but are implemented in different functions.This could be the root of issues like this. In the future, it would be better to have a common function for outputtingthe WAL summary file that both can use. It's entirely possible that, with some refactoring, more code could be shared. I tried to make all of the blkreftable stuff reusable, but I didn't pay as much attention to synchronizing up the various users of it. However, the fact that pg_walsummary is frontend code and pg_wal_summary_contents() is backend code does make it hard to get perfect reuse. I think if you go through pg_wal_summary_contents(), you'll find that almost every line of that function contains something backend-specific. -- Robert Haas EDB: http://www.enterprisedb.com
On 2024/07/03 22:42, Robert Haas wrote: > On Wed, Jul 3, 2024 at 5:34 AM Fujii Masao <masao.fujii@oss.nttdata.com> wrote: >> pg_wal_summary_contents() seems to miss the summary information with "limit" that pg_walsummary reports. This appearsto be a bug. The attached patch fixes this. > > Oops. It looks like pg_wal_summary_contents() forgets to emit the > limit block when that's the only data for a particular relation fork. > And maybe you can make it emit the limit block multiple times if the > list of block numbers is long enough. > > Thanks for the patch. I think you can commit and back-patch this, but > I don't think the commit message is quite right, because it's not like > this code just NEVER executes where it is located currently. Or am I > missing something? Yes, so I updated the commit message. I borrowed your description and used it in the message. Attached is the revised versionof the patch. If there are no objections, I will commit and backpatch it. Regards, -- Fujii Masao Advanced Computing Technology Center Research and Development Headquarters NTT DATA CORPORATION
Вложения
On Thu, Jul 4, 2024 at 6:16 AM Fujii Masao <masao.fujii@oss.nttdata.com> wrote: > Yes, so I updated the commit message. I borrowed your description and used it in the message. Attached is the revised versionof the patch. > > If there are no objections, I will commit and backpatch it. +1. Maybe change "Fix bugs in pg_wal_summary_contents()" to "Fix limit block handling in pg_wal_summary_contents()". -- Robert Haas EDB: http://www.enterprisedb.com
On 2024/07/08 22:50, Robert Haas wrote: > On Thu, Jul 4, 2024 at 6:16 AM Fujii Masao <masao.fujii@oss.nttdata.com> wrote: >> Yes, so I updated the commit message. I borrowed your description and used it in the message. Attached is the revisedversion of the patch. >> >> If there are no objections, I will commit and backpatch it. > > +1. Maybe change "Fix bugs in pg_wal_summary_contents()" to "Fix limit > block handling in pg_wal_summary_contents()". Thanks! I've pushed the patch and used your wording in the commit message. Regards, -- Fujii Masao Advanced Computing Technology Center Research and Development Headquarters NTT DATA CORPORATION