Re: Missing LWLock protection in pgstat_reset_replslot()

Hi!

b36fbd9f8d message says that inconsistency may still be possible because
statistics are not completely consistent for a single scan of
pg_stat_replication_slots under concurrent replication slot drop or
creation activity.

Seems there is a reproduction of such a case via isolation test.
Please see the repslot_stat.spec attached.

In an build with asserts performing such a test will result in a
crash in the checkpointer during server shutdown.

Please see postmaster.log and bt.txt. There are
wal_level = logical
max_replication_slots = 4
logical_decoding_work_mem = 64kB
autovacuum_naptime = 1d
log_min_messages = DEBUG
in the postgresql.conf.

The assert occures due to this stats entry:
(gdb) p *ps
$1 = {key = {kind = PGSTAT_KIND_REPLSLOT, dboid = 0, objoid = 0}, dropped = true, refcount = {value = 0}, generation = {value = 1}, body = 1099512025088}

Also the commit message says that such an issue should unlikely be a problem in practice.
So i doubt whether this should be treated as a bug and to be fixed?

Would be glad to figure this out.

With the best regards,

--
Anton A. Melnikov
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company