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