On 20.01.21 18:31, Emanuel Gonzalez wrote:
The problem can be generated by the number of processes?

number of concurrent processes trying to write to the bayes DB at the same
time.

journalling may help a bit, but it makes no sense to parse more mail within
one CPU at the same time.


# Server CPU

cpu family      : 6
model           : 60
model name      : Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz

4 cores, 8 threads. provided you only have one CPU.

I'd set max-children to 4 and not set min-children,min-spare and max-spare
at all.

... on some systems I disable HT CPUs by disabling in /etc/sysfs.conf:

devices/system/cpu/cpu4/online = 0
devices/system/cpu/cpu5/online = 0
devices/system/cpu/cpu6/online = 0
devices/system/cpu/cpu7/online = 0

I think since spectre/meltdown it's a good idea, and some systems reported
high dummy CPU usage when those were enabled.

# SpamAssassin

SPAMDOPTIONS="-u spamd --min-children=30 --max-children=80 --min-spare=25 
--max-spare=80 --timeout-child=60 --max-conn-per-child=150

ohh!  too many processes.  I don't recommend more spamd processes than e.g.
2x number of CPUs. maybe even less.
It does not make sense to run too many processes in parallel.

If you process too much mail, you could store bayes database in SQL or
redis. However, first lower amount of processes.

--
Matus UHLAR - fantomas, [email protected] ; http://www.fantomas.sk/
Warning: I wish NOT to receive e-mail advertising to this address.
Varovanie: na tuto adresu chcem NEDOSTAVAT akukolvek reklamnu postu.
10 GOTO 10 : REM (C) Bill Gates 1998, All Rights Reserved!

Reply via email to