From 2f6501fa3c54bbe4568e3bcccd9a60d26a46b5ee Mon Sep 17 00:00:00 2001 From: Andres Freund Date: Mon, 14 Feb 2022 16:44:28 -0800 Subject: Move replication slot release to before_shmem_exit(). Previously, replication slots were released in ProcKill() on error, resulting in reporting replication slot drop of ephemeral slots after the stats subsystem was already shut down. To fix this problem, move replication slot release to a before_shmem_exit() hook that is called before the stats collector shuts down. There wasn't really a good reason for the slot handling to be in ProcKill() anyway. Patch by Masahiko Sawada, with very minor polishing by me. I, Andres, wrote a test for dropping slots during process exit, but there may be some OS dependent issues around the number of times FATAL error messages are displayed due to a still debated libpq issue. So that test will be committed separately / later. Reviewed-By: Kyotaro Horiguchi Reviewed-By: Andres Freund Author: Masahiko Sawada Discussion: https://postgr.es/m/CAD21AoDAeEpAbZEyYJsPZJUmSPaRicVSBObaL7sPaofnKz+9zg@mail.gmail.com --- src/backend/tcop/postgres.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src/backend/tcop/postgres.c') diff --git a/src/backend/tcop/postgres.c b/src/backend/tcop/postgres.c index fda2e9360e2..38d8b97894c 100644 --- a/src/backend/tcop/postgres.c +++ b/src/backend/tcop/postgres.c @@ -4261,8 +4261,8 @@ PostgresMain(const char *dbname, const char *username) * We can't release replication slots inside AbortTransaction() as we * need to be able to start and abort transactions while having a slot * acquired. But we never need to hold them across top level errors, - * so releasing here is fine. There's another cleanup in ProcKill() - * ensuring we'll correctly cleanup on FATAL errors as well. + * so releasing here is fine. There also is a before_shmem_exit() + * callback ensuring correct cleanup on FATAL errors. */ if (MyReplicationSlot != NULL) ReplicationSlotRelease(); -- cgit v1.2.3