Hi,
When a hot standby is restarted in a state where subtransactions have
overflowed, it may become inaccessible:
$ psql: error: connection to server at "localhost" (::1), port 5433
failed: FATAL: the database system is not yet accepting connections
DETAIL: Consistent recovery state has not been yet reached.
However, the log message that indicates the cause of this issue seems to
be only output at the DEBUG1 level:
elog(DEBUG1,
"recovery snapshot waiting for non-overflowed snapshot or "
"until oldest active xid on standby is at least %u (now %u)",
standbySnapshotPendingXmin,
running->oldestRunningXid);
I believe this message would be useful not only for developers but also
for users.
How about changing the log level from DEBUG1 to NOTICE or else?
Background:
One of our customers recently encountered an issue where the hot standby
became inaccessible after a restart.
The issue resolved itself after some time and I suspect it was caused by
a subtransaction overflow.
If the log level had been higher one, it would have been easier to
diagnose the problem.
..Even if it was a NOTICE, it may be difficult to notice the cause if
the log_min_message is set to default WARNING, but well, it seems a
higher log level is better than DEBUG1.
I would appreciate your thoughts.
--
Regards,
--
Atsushi Torikoshi
Seconded from NTT DATA GROUP CORPORATION to SRA OSS K.K.
diff --git a/src/backend/storage/ipc/procarray.c b/src/backend/storage/ipc/procarray.c
index 2e54c11f88..20c505add7 100644
--- a/src/backend/storage/ipc/procarray.c
+++ b/src/backend/storage/ipc/procarray.c
@@ -1125,7 +1125,7 @@ ProcArrayApplyRecoveryInfo(RunningTransactions running)
"recovery snapshots are now enabled");
}
else
- elog(DEBUG1,
+ elog(NOTICE,
"recovery snapshot waiting for non-overflowed snapshot or "
"until oldest active xid on standby is at least %u (now %u)",
standbySnapshotPendingXmin,
@@ -1303,7 +1303,7 @@ ProcArrayApplyRecoveryInfo(RunningTransactions running)
if (standbyState == STANDBY_SNAPSHOT_READY)
elog(DEBUG1, "recovery snapshots are now enabled");
else
- elog(DEBUG1,
+ elog(NOTICE,
"recovery snapshot waiting for non-overflowed snapshot or "
"until oldest active xid on standby is at least %u (now %u)",
standbySnapshotPendingXmin,
