Busy Zabbix Poller - How to get rid of it ?

[Maverick1980]

Hello,

we are monitoring our vmware environment with zabbix.
(Version 2.4 of zabbix).

We had to configure more VMware Pollers because it shows warning of busy vmware collector. So we confiured the vmware pollers up to 50.
We also configred StartPollers to 50.

Vmware Pollers are now working fine. But zabbix now shows "Busy Zabbix Poller Warning". It is always at 100%! (see attachment)

If i look to the queues on zabbix, there are also a lot of items within 1 minute queue for simple check and zabbix agent.

What can we do to get a better performance for it ?

Thanks for your answer,

Mav

[LenR]

What does the Required server performance, new values per second show? What is your timeout value? How is you database performance? How many cores on the server? How is it's load?

[Maverick1980]

Hello,

thanks for your answer.

This is the server configuration:

RedHat Enterprise 6
8 vCPU
16 GB RAM
DB on Oracle

Number of hosts (enabled/disabled/templates) 2753 2731 / 0 / 22
Number of items (enabled/disabled/not supported) 153753 150129 / 0 / 3624
Number of triggers (enabled/disabled [problem/ok]) 7808 7808 / 0 [40 / 7768]
Required server performance, new values per second 2170.44

Where do i have to set Timeout value?

Thanks for your answer,

Mav

[LenR]

TImeout is set in zabbix_server.conf, the default is 3. I'd try bumping pollers to 200 and see what happens. Once you get under 100% busy, you can reduce pollers.

[tchjts1]

Instead of taking a reactive approach, take a proactive one.

Have a look at my post here, the last paragraph and the graphs that are with it:

https://www.zabbix.com/forum/showthread.php?t=41219

Those graphs are already part of your Linux template as a default. Check your pollers, trappers, cache usage. All those things combined affect your performance.

Checking those an a regular basis and making necessary adjustments as you add in items/triggers/hosts is going to help your Zabbix setup run more smoothly and efficiently.