rspamd processes crash

[chri2]

Describe the bug

rspamd crashes regularly:

2023-10-17 13:27:42 #26134(normal) rspamd_crash_sig_handler: caught fatal signal 11(Speicherzugriffsfehler), pid: 26134, trace: 
2023-10-17 13:27:42 #1368(main) <316489>; main; rspamd_check_termination_clause: normal process 26134 terminated abnormally with exit code 0 by signal: Speicherzugriffsfehler but NOT created core file (throttled=no); core file limits: 0 current, -1 max
2023-10-17 13:27:43 #26140(normal) rspamd_crash_sig_handler: caught fatal signal 11(Speicherzugriffsfehler), pid: 26140, trace: 
2023-10-17 13:27:43 #1368(main) <316489>; main; rspamd_check_termination_clause: normal process 26140 terminated abnormally with exit code 0 by signal: Speicherzugriffsfehler but NOT created core file (throttled=no); core file limits: 0 current, -1 max
2023-10-17 13:27:45 #28514(normal) rspamd_crash_sig_handler: caught fatal signal 11(Speicherzugriffsfehler), pid: 28514, trace: 
2023-10-17 13:27:45 #1368(main) <316489>; main; rspamd_check_termination_clause: normal process 28514 terminated abnormally with exit code 0 by signal: Speicherzugriffsfehler but NOT created core file (throttled=no); core file limits: 0 current, -1 max

Context

• Hardware: RockPro64 + 2x SATA SSD as software raid level 1
• YunoHost version: 11.2.5
• I have access to my server: Through SSH | through the webadmin
• Are you in a special context or did you perform some particular tweaking on your YunoHost instance?: yes
  • If yes, please explain: multiple domains for email, added aliases as described here

To reproduce

• Look at your logfile. No?
• Look at the logfile of your Rock64Pro yunohost. No?
• Look at the logfile of your Rock64Pro SATA SSDs software raid 1 yunohost. No?

Can't say so far. I can increase the rate of the error significantly by running rspamc learn_spam /var/mail/<user>/.junk/cur on a big junk mailbox.

Expected behavior

No crashes at all.

Logs

On systems affected

for P in $(grep Speicherzugriffsfehler /var/log/rspamd/rspamd.log | perl -pe's/.*#(\d+)\(.*/$1/' | sort | uniq); do grep "#$P(" /var/log/rspamd/rspamd.log; done

should show the logs of the processes that crashed.

[chri2]

I tried to dig a bit more and switched my system wide locale from de_DE.UTF-8 to C.UTF-8 to get the rspamd logfile in pure english without german mixed in. Restarted rspamd using systemctl.

Not reproducible.

Switched the locale back to de_DE.UTF-8, restarted rspamd and reproduced the bahviour.

To get my ynh working with rspamd I switched back to C.UTF-8, restarted rspamd and got the same errors containing german again. WTF.

After I restarted the daemon a few times and tried after each time to learn more ham/spam from existing folders it suddenly started working without crashes again.

I guess someone needs to look at the problem on their host to compare results.

[chri2]

This might be related: https://github.com/rspamd/rspamd/issues/3849

[alexAubin]

Closing because it's unclear, and rspamd is to become an app on Bookworm / Yunohost 12.x