zabbix poller process more than 75% busy

[ajay]

I am using zabbix v 4.0.17 at a very large scale.
hosts configure= 505
items: almost 100 for each hosts (both active and passive)

while digging in the above mentioned alert i found out I have not configured any value to startpoller flag in zabbix_server.conf
by default it takes values a 5.
(ref: https://blog.zabbix.com/monitoring-how-busy-zabbix-processes-are/457/)

as verified from below command:
---
grep -P 'server: poller #' /proc/*/cmdline
/proc/3722/cmdline:/usr/sbin/zabbix_server: poller #1 got 260 values in 4.955364 sec, getting values
/proc/3723/cmdline:/usr/sbin/zabbix_server: poller #2 got 340 values in 4.982723 sec, getting values
/proc/3724/cmdline:/usr/sbin/zabbix_server: poller #3 got 308 values in 4.967018 sec, getting values
/proc/3725/cmdline:/usr/sbin/zabbix_server: poller #4 got 357 values in 5.033239 sec, getting values
/proc/3726/cmdline:/usr/sbin/zabbix_server: poller #5 got 329 values in 4.988694 sec, getting values
---

given that the above scenario, what should I put the value of this flag, is there any formula or mechanism to determine the ideal or good value for this flag.

thanks.

[kola]

I am running into this issue also. Mine seems to be happening on the hour. I have a Zabbix server and 3 Zabbix proxies in remote branches and i noticed all 4 servers would alert with poller processes more than 75% busy. Any suggestions would be appreciated.

[Atsushi]

The poller process is used to get the value of the item that is actively retrieved from the Zabbix server. If you have a lot of hosts or items to monitor, it would be too late to process 5 processes in parallel.
Use the value of NVPS on the dashboard as a reference to estimate how many processes should be executed at the same time. For example, raise the value of Start Pollers a little and restart the Zabbix server and see how it works.
ex. 5 to 10

[kola]

The poller process is used to get the value of the item that is actively retrieved from the Zabbix server. If you have a lot of hosts or items to monitor, it would be too late to process 5 processes in parallel.
Use the value of NVPS on the dashboard as a reference to estimate how many processes should be executed at the same time. For example, raise the value of Start Pollers a little and restart the Zabbix server and see how it works.
ex. 5 to 10

Thanks for your comment. Here is a screenshot of my values per second maximum reached about 595. how would i determine the Start Pollers from this value?

[Atsushi]

The appropriate value cannot be derived by simple calculations, as the time taken for each item will vary. Repeat changing the value of StartPollers and checking the busy rate, and try adjusting it so that it is 70% or less.

[kola]

So i changed StartPollers to 10 on the server and the proxies. That seemed to work. I noticed values per second is now peaking at about 425 now not as much as 500.
Will keep an eye on things.
Thanks

[pohlepihp]

The poller process is used to get the value of the item that is actively retrieved from the Zabbix server. If you have a lot of hosts or items to monitor, it would be too late to process 5 processes in parallel.
Use the value of NVPS on the dashboard as a reference to estimate tellthebell how many processes should be executed at the same time. For example, raise the value of Start Pollers a little and restart the Zabbix server and see how it works.
ex. 5 to 10

Thank you so much for sharing such a useful information. I will definitely share this with others.

[fatih gur]

Hello Everyone ,

I am using zabbix v 5.0
Host configured = 85
items: almost 80-90 for each hosts (passive)

I am having alert ; Zabbix trapper processes more than 75% busy and Zabbix discoverer processes more than %75 busy then It caused to zabbix is not running even systemctl status zabbix-server is" active ".After zabbix-server.service restart , In dashborad zabbix server status is running but after for a while status changed as "not running"
I attached output of zabbix systemctl and zabbix-server_log
Anyone can help me ?

ubuntu@zabbix-server:~$ sudo tail -f /var/log/zabbix/zabbix_server.log
25399:20201103:072229.891 server #33 started [alerter #3]
25398:20201103:072229.891 server #32 started [alerter #2]
25400:20201103:072229.891 server #34 started [preprocessing manager #1]
25404:20201103:072229.894 server #38 started [lld manager #1]
25409:20201103:072229.911 server #39 started [lld worker #1]
25410:20201103:072229.924 server #40 started [lld worker #2]
25412:20201103:072229.951 server #41 started [alert syncer #1]
25401:20201103:072229.952 server #35 started [preprocessing worker #1]
25402:20201103:072229.962 server #36 started [preprocessing worker #2]
25403:20201103:072229.976 server #37 started [preprocessing worker #3]

[avidec87]

Hi Faith,

I hope you issue should be fixed by now but just for other reference, you need to monitoring your discovery period defined inside the discover. It should be 1 hour or more. Generally if somebody changes the discover timeperiod for a shorter time period for testing such as minutes or seconds and forget to update then it will create issue which you are seeing in your environment.