Too many processes on zabbixserver

[hernan]

Hello Team,

Problem description:

I received the following error message on the Zabbix Server:

"Too many processes on zabbixserver"

I checked the process running on the machine and I found that the Zabbix Server has 1397 process running:


~$ ps -ef | awk '{ print $8 }' | sort -n | uniq -c | sort -n | tail -5
3 lightdm
5 /usr/sbin/zabbix_agentd:
6 /sbin/getty
11 /usr/sbin/apache2
1397 /usr/sbin/zabbix_server:
~$

Can you help me to solve this issue?

Thanks in advance.

Regards.

[LenR]

The "too many processes" trigger is a sample in the default template, it's only a starting point. Depending on the number of pollers and other things in your zabbix_server.conf, you will expect to have a LOT of processes on your zabbix server.

If you want the alert, clone the trigger on this box, set the limit to 1450 or so, then disable the standard trigger; or just disable the original trigger.

Linux will have a number of procs per core by default, I think it's 3, so if you have a 24 core box, that's 72 procs just for the kernel.

[hernan]

The "too many processes" trigger is a sample in the default template, it's only a starting point. Depending on the number of pollers and other things in your zabbix_server.conf, you will expect to have a LOT of processes on your zabbix server.

If you want the alert, clone the trigger on this box, set the limit to 1450 or so, then disable the standard trigger; or just disable the original trigger.

Linux will have a number of procs per core by default, I think it's 3, so if you have a 24 core box, that's 72 procs just for the kernel.

Thank you LenR !!! I will follow your suggestion on my environment.

Regards.

[volter]

1400 zabbix_server processes?

[LenR]

Yea, second thought, that is a LOT, my server is about 350 and about that on several proxies. I guess if we tried to run ALL of it on the server, it could add up, but it probably wouldn't run right.

[volter]

Take a good look at your configuration. You probably configured way too many workers of some kind. Also keep in mind that each of them has a permanent database connection!

[kloczek]

Take a good look at your configuration. You probably configured way too many workers of some kind. Also, keep in mind that each of them has a permanent database connection!

More likely he increased the number of pollers to with growing number of hosts and passive items to monitor.
If it is the case he needs to switch from passive monitoring to active one.
Passive monitoring does not scale well.
The second thing is adding the proxy. Even with only one proxy is possible to minimize losses of monitoring data during the server restart.