Thanks a lot for your input!
You wrote:
Lars Bjørndal writes:
You've got some kind of logging problem with writing to the persistent storage for the log, whatever that is (defaults to several files in $var/dir/logs, but can be configured very flexibly using the stdlib logging module).
I've installed mailman3 following the guide at https://docs.mailman3.org/en/latest/install/virtualenv.html. So Mailma's log files is in /opt/mailman3/mm/var/log. But the log I've provided, is, as you say, from systemd/journalctl.
A bunch of Mailman processes are raising exceptions when trying to log that they're shutting down. This could result in a deadlock, which is why systemd terminates the processes with SIGABRT.
There's also a message about reentrancy (I think only one per process, but Mailman 3 spawns lots of processes). It might be a bug in logging (not Mailman code) but I think that's some exception handler that's trying to log the logging exception. If so, it's not the root problem and should not recur if you fix the basic problem.
- What, if anything, is in Mailman's own logs? This appears to be systemd's log from journalctl.
In /opt/mailman3/mm/var/logs/error.log I have:
[2026-07-20 08:57:39 +0200] [2594] [INFO] Handling signal: int
[2026-07-20 08:57:41 +0200] [2594] [ERROR] Worker (pid:2596) was sent code 131!
[2026-07-20 08:57:42 +0200] [2594] [INFO] Shutting down: Master
[2026-07-20 09:02:00 +0200] [2510] [INFO] Starting gunicorn 23.0.0
[2026-07-20 09:02:00 +0200] [2510] [INFO] Listening at: http://127.0.0.1:8000 (2510)
[2026-07-20 09:02:00 +0200] [2510] [INFO] Using worker: sync
[2026-07-20 09:02:00 +0200] [2538] [INFO] Booting worker with pid: 2538
- Do you have any settings for logging in mailman.cfg?
No, nothing.
- If your logging handler is a file-based handler, maybe the log partition is full.
Both /opt and /var is on the same file system with lot of space.
- Maybe there is some security restriction (selinux?) on the logging channel, whatever it is. Or a permissions restriction on the file system so Mailman can't write.
I don't think so, because Mailman writes to it's logs in /opt/mailman3/mm/var/log. I don't use selinux.
I also noticed that PostgreSQL gets shut down, but there's no log that it gets restarted. I think Mailman should then shutdown with an error that it can't talk to the database, but maybe it retries a few times.
From the log I can see that PostgeSQL is started before mailman tries to start.
Lars