Zabbix monitoring environment on-line deployment after a period of time, the alarm suddenly appeared "Zabbix poller processes more than 75% busy"
In fact, there are many kinds of monitoring alarm Zabbix, more common than memory exhaustion, network is not through, Io is too slow and this "Zabbix poller processes more than 75% busy". The beginning of the time because it does not affect the use of a few minutes to solve on their own don't care. Then as the database grows, Zabbix consumes more and more memory, Poller processes (polling) starts every day busy.
Finally, finding a solution to this problem is simple!
You can increase the number of processes that are initialized at Zabbix server startup, but doing so directly increases the amount of polling load that can be done with sufficient memory configuration.
Specifically edit the configuration file for Zabbix server/etc/zabbix/zabbix_server.conf, locate the paragraph that configures startpollers:
# # Option:startpollers
# Number of pre-forked instances of pollers.
#
# Mandatory:no
# range:0-1000
# Default:
# startpollers=5
Cancel the # comment before startpollers, modify 5 to 10 or greater "due to the online machine memory 64G, I change here to 30 or 60"
After modifying, restart Zabbix_server
#pkill-9 Zabbix_server
#/usr/local/zabbix/sbin/zabbix-server
After a while, we found no such warning in the trigger.
Of course, we can also write a script at the rated time to restart the zabbix_server to reduce the load
Here is the script/root/zabbix-restart.sh
#!/bin/bash
/usr/bin/pkill Zabbix_server
/usr/local/zabbix/sbin/zabbix_server
And then crontab to do the planning tasks.
0 3 * * */bin/bash-x/root/zabbix-restart.sh >/dev/null 2>&1
Zabbix monitoring field appears "Zabbix Poller processes more than 75% busy" alarm