Zabbix server sizing

[gcarmichael]

Good morning,
I am in the process of trying to build out a Zabbix server to monitor my 300 some odd switches and routers combined total. I keep having the issue of it complaining about Value CacheSize and CacheSize being 75+% used. I can not seem to find how to size this thing other than random guessing. Is there a matrix or some other doc some where that tells you what you should need for resources for the server and the config? Im doing this as SolarWinds is going to be discontinuing our particular product and Im trying to get ahead of the game. Thanks for any help you can provide.

[kyus]

It really depends on how many metrics you are going to monitor.
For comparison, in one server I have about 38k items, 700 hosts, and around 530 new values per second. Configuration cache (CacheSize in the conf file) is set to 256M and the average usage is 35%. ValueCacheSize is at 32M and around 10% average usage.

I would suggest that you increase it to a somewhat high value, accordingly to the amount of items and triggers.
Then add your hosts, attach templates, let all items be discovered, check the "Zabbix server health" dashboard and verify the usage of caches and pollers.
After that you can adjust everything based on the usage, usually having values between 20-60% is ideal.

If either CacheSize or ValueCacheSize get full, your server will crash.

Also, HistoryCacheSize and TrendCacheSize are very important because they're used to store data temporarily.

You'll probably want to increase the amount of pollers aswell, possibly:
StartPingers
StartSNMPPollers
StartPollers
StartPollersUnreachable

Lastly, It's important to tune your database, accordingly to the resources available in the VM.

https://www.zabbix.com/documentation/current/en/manual/installation/requirements

[PavelZ]

What version are you using?
First, I suggest checking this problem https://support.zabbix.com/browse/ZBX-25994

[gcarmichael]

7.2.3

I got it figured out for the moment. Seems to be happy running. I found another article that had a similar deployment and that person had included their CacheSize settings for all of them.