Am 06.02.24 um 17:09 schrieb Christian Mack (christian.m...@uni-konstanz.de):
Hello
Am 06.02.24 um 13:41 schrieb Frank Richter
(frank.rich...@hrz.tu-chemnitz.de):
Hi,
we’ve set SxVMemLimit = 768;
in /etc/sogo/sogo.conf
But it seems that this big value isn’t sufficient, we see this several
times a day in sogo.log:
Feb 06 04:47:06 sogod [1280936]: 2003:f3:1708:b0ca:f955:2664:d05:4987
"GET /SOGo/so/dadel/Mail/0/view HTTP/1.1" 200 427415/0 1.657 - - 39M - 20
Feb 06 04:47:06 sogod [1280936]: |SOGo| terminating app, vMem size limit
(768 MB) has been reached (currently 771 MB)
ax_Main;"
^
[...]
Feb 06 04:47:06 sogod [1280926]: <0x0x5606e7615030[WOWatchDogChild]>
child 1280936 exited
What is advisable? We don’t like to raise SxVMemLimit even more …
They will always happen.
But such worker/child termination is only happening *after* a request was
processed.
Therefore they are usually harmless.
The only drawback is the time needed to start a new worker/child.
While this is happening, you have less worker available.
Therefore you only have to keep an eye on how often this happens.
As long as it only happens a couple of times per hour, you are save.
Thanks, Christian, for clarifying this.
I’ll look into memory consumption of the SOGo processes, and maybe revert to
512 MB.
BTW:
We use SxVMemLimit = 512;
with 350 worker
With approx. 20,000 accounts => approx. 1.2 million requests per day
=> approx. 600 of the above terminating errors per day
We have 9,500 users , use 2 servers with each 16 GB RAM and 50 SOGo
workers, ~ 500,000 requests per work day on each server.
Frank
--
Frank Richter, Chemnitz University of Technology, Germany