Zabbix agent is not available - up and down repeatedly

[buttershave99]

The last two days we've been seeing a lot of alerts saying the Zabbix agent is not available. Shortly after we receive the alert that it's back up. The "downtime" reported is anywhere from 30s to 5min. Zabbix version is 6.4.1.

We have 40+ Windows Servers running the Zabbix agent. I changed the agent to use IP address instead of DNS but it made no difference. The Zabbix server was reporting 95% memory usage so I restarted it but so far we're still getting alerts. I checked the time on some of the servers and they look the same compared to Zabbix.

I inherited this setup and don't have a lot of experience yet. Appreciate any help. Thanks.

[buttershave99]

Also seeing this alert:
Utilization of poller processes over 75%

[tim.mooney]

Are all the components (the database backend, the Zabbix server, and the Zabbix web front-end) running on the same system? Are you using traditional packages, or containers, or how is stuff installed?

If you're seeing an alert that your poller processes are more than 75% busy, then you probably want to slightly increase the number of poller proceses the Zabbix server uses. Assuming you're using traditional packages for your install, look at the setting for "StartPollers" in /etc/zabbix/zabbix_server.conf. You want to increase that, but don't just set it to a huge value.

The problem is that increasing the # of pollers will likely increase resource usage, including memory usage, so you may need to allocate more resources to this system (if that's possible). Again, not knowing anything about your server environment makes it difficult to offer specific suggestions.

It's also possible that you don't really need more pollers, and the pollers being busy is a symptom of some other problem, but not knowing more about your environment, the best thing to try is to slightly increase StartPollers, and then see if other fixes will be needed.