| keithzg | |
| Nov 12 2024, 6:36 PM |
| F175628: image.png | |
| Nov 12 2024, 6:36 PM |
Perhaps it's because now it's actually being used by more than me, perhaps it's due to changes with T189: Upgrading kloud (the OOM killer is a lot more aggressive by default in newer *buntu versions), but this is not ideal:
Certainly there is uhh not that much RAM:
total used free shared buff/cache available Mem: 957Mi 204Mi 57Mi 14Mi 695Mi 574Mi Swap: 1.5Gi 104Mi 1.4Gi
but, still...
| Status | Assigned | Task | ||
|---|---|---|---|---|
| Restricted Maniphest Task | ||||
| Open | keithzg | T189 Upgrading kloud | ||
| Open | keithzg | T202 Make kloud a bit more robust |
https://mariadb.com/kb/en/systemd/#configuring-the-systemd-service led me to doing this:
[System] OOMScoreAdjust=-999
That's literally 1 before it would just never kill any processes owned by the database service, which seems like a good moderate choice!
Rebooting now . . .
The conf got loaded on reboot, so in theory it's all good, but I should perhaps give it some more swap or RAM (and probably disk space?)
Looking at the logs, I got the section name wrong. So I've changed it to:
[Service] OOMScoreAdjust=-1000
But then, I tried rebooting the server and the webserver itself got OOM killed shortly after boot!
1GB is a paltry amount of RAM, so I've bitten the bullet and set to oay the extra mere $6 a month for 2GB RAM on Kloud. Maybe that'll be enough to avoid this problem altogether...