Behavior when one of the servers if offline

[hover]

Hello,

Please look at the picture. The graphs of all servers have been "patchy" and consisted of disconnected fragments during the period when one of the servers lacked zabbix_agentd, and instantly regained normal shape once that process was started.

My setup includes 9 running copies of zabbix_server (while only 7 servers are monitored). Each client runs 6 copies of zabbix_agentd. Could this strange interdependency of monitored data be explained by wrong process numbers choice?

[Alexei]

It seems that ZABBIX server was unable to connect to the host (one without zabbix_agentd running) causing 5 seconds delays. It is strange why the host didn't reject the connections immediately.

I would suggest to switch to active checks (latest 1.1 alphas). This will avoid cases like this. The active checks will also significantly (2x) improve ZABBIX server performance.