Incomplete Graphs / Missing Data?

[selin]

Attached is a PNG of a graph generated from Zabbix. Notice that there is a gap in the middle of the chart.

Has anyone seen this behavior? If so, what causes it?

The machine in question was independently verified to be available during the same time period via ICMP ping and a Nagios check on the MSSQL port.

As far as I can tell, when this happens, all of the hosts monitored by Zabbix are affected. The network appears to be fine. The Zabbix logs give me a "Timeout" error at the time. It happens more often when SNMP sampling is occurring.

Does Zabbix spawn processes, or do each of the X number of server processes (configured in /etc/zabbix/zabbix_server.conf) handle the traffic? Should I switch the system to use xinetd and generate a new thread (terminology ?) for each connection?

So far, Zabbix has been working well - except for this one feature. We have decided to configure a clustered monitoring environment to ensure that Zabbix is always up and running...

Based on my experience so far, I am certain that I am being thick-headed. I have attempted to search the forums for an answer and have not found anything identical to this issue.

[Alexei]

Hmm... Your processor load hits 10, I'm not suprised you start getting timeouts from ZABBIX agents.

[Nate Bell]

I have noticed that I sometimes get gaps in my data when an item (probably always a custom one) is taking longer than Zabbix allows to return data. Your gaps in data may be caused by the host being unable to return the requested data in the timelimit Zabbix sets (by default 3 seconds). I've managed to clear this up by reducing the load on the server, or setting the amount of time Zabbix waits for an item to return to a higher value in the .conf files (don't know if it's a 1.1 option only).

Nate