Influx of false alerts with non-default zabbix_server.conf

[ILIV]

This is a somewhat powerful server with plenty of resources.

Intel(R) Xeon(R) CPU L5430 @ 2.66GHz x 8 cores

$ free -g
             total       used       free     shared    buffers     cached
Mem:            23         19          3          0          0         18
-/+ buffers/cache:          1         22
Swap:           23          0         23

Database PostgreSQL server is running on the same host.

Total NVPS is only 32.87

I've adjusted configuration parameters in zabbix_server.conf to account for available resources but the effect was surprisingly the opposite of what I expected. I started to get a lot of false SSH down and other alerts. I had tried a lot of things but all those problems disappeared as soon as I reverted zabbix_server.conf to its defaults.

So, here's the customized zabbix_server.conf version that caused the problems:

StartPollers=40
StartPollersUnreachable=10
StartTrappers=30
StartPingers=20
StartDiscoverers=10
CacheSize=1024M
CacheUpdateFrequency=240
HistoryCacheSize=256M
TrendCacheSize=128M
HistoryTextCacheSize=128M
Timeout=20
UnreachablePeriod=90
LogSlowQueries=3000
CacheSize=1024M
CacheUpdateFrequency=240
HistoryCacheSize=256M
TrendCacheSize=128M
HistoryTextCacheSize=128M
MaxHousekeeperDelete=5000

I was wondering if anyone can see anything wrong with it. It seems okay to me, but in practice it didn't work out well. Like at all.

[ILIV]

Surprisingly, the Timeout parameter was the culprit. It doesn't make much sense to me but the default for this parameter solved the problem. Whereas the custom value of 20 caused all the problems.