Windows service randomly stops with no indication why

On Fri, Mar 8, 2024 at 9:17 AM Jay Madren <jay@jaymadren.com> wrote:

Running PostgreSQL 15.6 on Windows Server 2022. The database service
randomly just stops and the Windows Service auto-restart options don't kick
in. The stop is unexpected (not a controlled shut down) because after
restarting the service the postgresql log states that the database system
was interrupted, not properly shut down, and automatic recovery in
progress. There's nothing to indicate exactly when it went down, other than
the timestamp of the last log entry before restarting it.

There's also nothing I could find in the Windows Event Logs other than
when the service is started each time.

This is occurring at random times, day or night, around 0 to 2 times each
day.

Any clue as to what's happening or how to investigate this further?

What is log_min_messages set to? Increasing it might shed some light.

Go to "Services", find Postgres, and try a manual restart, see what
happens. Then check the recovery tab and see what it is supposed to do on
failures - you are probably at the "Do nothing" count limit, hence the no
auto restart.

If you can manually duplicate the failure to restart, try increasing the
logging, removing extensions, simplifying the postgresql.conf, etc, until
it starts working again. I suspect it will restart fine, though. Maybe
check anti-virus software logs?

Cheers,
Greg

On 08/03/2024 14:17, Jay Madren wrote:

Running PostgreSQL 15.6 on Windows Server 2022. The database service
randomly just stops and the Windows Service auto-restart options don't
kick in. The stop is unexpected (not a controlled shut down) because
after restarting the service the postgresql log states that
the database system was interrupted, not properly shut down, and
automatic recovery in progress. There's nothing to indicate exactly
when it went down, other than the timestamp of the last log entry
before restarting it.

There's also nothing I could find in the Windows Event Logs other than
when the service is started each time.

This is occurring at random times, day or night, around 0 to 2 times
each day.

Any clue as to what's happening or how to investigate this further?

I haven't used Windows in a very long time, but from (dim and distant)
memory PostgreSQL on Windows maintains its own log files, separate to
the Windows event log - you may find more clues there.

Ray.

--
Raymond O'Donnell // Galway // Ireland
ray@rodonnell.ie

On 3/8/24 06:17, Jay Madren wrote:

Running PostgreSQL 15.6 on Windows Server 2022. The database service
randomly just stops and the Windows Service auto-restart options don't
kick in. The stop is unexpected (not a controlled shut down) because
after restarting the service the postgresql log states that the database
system was interrupted, not properly shut down, and automatic recovery
in progress. There's nothing to indicate exactly when it went down,
other than the timestamp of the last log entry before restarting it.

I see this sort of thing on Windows I immediately suspect Anti-Virus
software. Verify whether you have AV software running and if it is that
it is scanning the Postgres files.

There's also nothing I could find in the Windows Event Logs other than
when the service is started each time.

This is occurring at random times, day or night, around 0 to 2 times
each day.

Any clue as to what's happening or how to investigate this further?

Thanks,
Jay

--
Adrian Klaver
adrian.klaver@aklaver.com

What is log_min_messages set to? Increasing it might shed some light.

It is not set (commented out). I assume the default is "warning". I will
set it to "info" and see if that reveals anything.

Then check the recovery tab and see what it is supposed to do on failures

- you are probably at the "Do nothing" count limit, hence the no auto
restart.

I had already changed the recovery options to restart on first and second,
then run a command script that starts it on the third count. None of them
appear to work, which means the shutdown appears to be "normal" to Windows.
Or maybe postgres doesn't set exit levels for Windows to act on.

I see this sort of thing on Windows I immediately suspect Anti-Virus

software.

This is what I suspect also. This server was recently "taken over" by a
third-party management firm that corporate dictated, and they have loaded
all kinds of monitoring and security crap on it. I have reached out to them
but not gotten a response. I will push harder on this.

Thanks,
Jay