I get a similar problem a few times a year.
Usually I just reboot mailcow stack and I’m good for a few months.
But it got more frequent recently. Maybe because memory requirements indeed changed lately.
That is I have 4 MB of RAM and requirements were lower than that not that long ago.
I noticed that today around 13h25 (1:25 PM) my host load jumped to the roof.
Total CPU usage and rspam container memory usage then went up and down hill non-stop.
When I checked command top around 19h tonight Rspamd was using 98-100% CPU.
Here are logs of the rspamd container from 13h00 until around 19h30 when I noticed the issue.
I so rebooted Mailcow stack and everything then settled calm.
N.B. Disregard the random-numbered rspamd container that can be seen in screenshot, I had made a mistake while trying to reboot it.
Any idea of what is happening?