Agent on Windows Server goes unavailable nightly

[aaron.axvig]

I have the 2.4 agent running on a Windows Server 2012 R2 virtual machine. Every night I get about 20 "Agent on sc is unreachable for 5 minutes" notifications. Where do I start my investigation?

My only clues so far is that this server runs Data Protection Manager and this time of night is when it is backing up a bunch of other servers. This would mean high network usage but CPU and memory are not taxed. Attached is a picture of the network graph overnight where you can see it dropping out.

[acropia]

This could have many causes.

The steps I would do, would be:

1. Have a look at the Zabbix Queue
You can find the Zabbix Queue in Zabbix Web / Administration / Queue. When an item can not be monitored by whatever reason, it will be added to the queue. Ideally the queue would be empty. But when monitoring hundreds/thousands of hosts there is a change some items will queue. But when this is excessive, measurements will get lost.

2. Server/Proxy logs
Check the logs of the server which is contacting the Zabbix Agent. In small environments this will be most likely the Zabbix Server. In distributed environments this also could be a Zabbix Proxy.

It helps when you raise the DebugLevel to 4. This can be set in the Zabbix Server config:

/etc/zabbix/zabbix_server.conf

Change the value of DebugLevel to 4, save the config and restart the Zabbix Server service. On my CentOS Linux 7.1 machine with systems this is done by:

systemctl restart zabbix-server

Now you can investigate the server logs at:

/var/log/zabbix/zabbix_server.log

Because these will grow extremely fast with DebugLevel 4, filter the logs by using tail and grep, substitute hostname by your actual hostname of the agent:

tail -f /var/log/zabbix/zabbix_server.log | grep -i hostname

I think with these steps you should be able to investigate your problem.