When walsender calls out to do_pg_stop_backup() (during base backups), it is not possible to terminate the process with a SIGTERM - it requires a SIGKILL. This can leave unkillable backends for example if archive_mode is on and archive_command is failing (or not set). A similar thing would happen in other cases if walsender calls out to something that would block (do_pg_start_backup() for example), but the stop one is easy to provoke.
ISTM one way to fix it is the attached, which is to have walsender set the "global" flags saying that we have received sigterm, which in turn causes the CHECK_FOR_INTERRUPTS() calls in the routines to properly exit the process. AFAICT it works fine. Any holes in this approach? Second, I wonder if we should add a SIGINT handler as well, that would make it possible to send a cancel signal. Given that the end result would be the same (at least if we want to keep with the "walsender is simple" path), I'm not sure it's necessary - but it would at least help those doing pg_cancel_backend()... thoughts? -- Magnus Hagander Me: http://www.hagander.net/ Work: http://www.redpill-linpro.com/
diff --git a/src/backend/replication/walsender.c b/src/backend/replication/walsender.c index 5d1c518..c8fd165 100644 --- a/src/backend/replication/walsender.c +++ b/src/backend/replication/walsender.c @@ -1302,6 +1302,13 @@ WalSndShutdownHandler(SIGNAL_ARGS) if (MyWalSnd) SetLatch(&MyWalSnd->latch); + /* + * Set the standard (non-walsender) state as well, so that we can + * abort things like do_pg_stop_backup(). + */ + InterruptPending = true; + ProcDiePending = true; + errno = save_errno; }
-- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
