1.3.5 server hogs cpu and log file if no zabbix agent is running on a monitored host

[Farzad FARID]

Hi,

I voluntarily added a new host to Zabbix Server, with about 20 items (mostly Zabbix agent type + one simple check: SSH), without installing the Zabbix agentd daemon on the monitored host.

Instantaneously, Zabbix server seems to go in a wild loop, trying to check the services hundreds of times per seconds, and spewing as much as 3800 lines/second in the log file:

 26131:20070419:165053 Getting value of [proc.num[]] from host [ora0] failed
 26131:20070419:165053 The value is not stored in database.
 26135:20070419:165053 Getting value of [system.cpu.load[,avg5]] from host [ora0] failed
 26135:20070419:165053 The value is not stored in database.
 26130:20070419:165053 Getting value of [system.swap.size[,free]] from host [ora0] failed
 26130:20070419:165053 The value is not stored in database.
 26134:20070419:165053 Getting value of [vm.memory.size[shared]] from host [ora0] failed
 26133:20070419:165053 Getting value of [vm.memory.size[free]] from host [ora0] failed
 26134:20070419:165053 The value is not stored in database.
 26133:20070419:165053 The value is not stored in database.
 26136:20070419:165053 Getting value of [vfs.dev.write[sdb,operations]] from host [ora0] failed
 26136:20070419:165053 The value is not stored in database.
...
etc.

The CPU load avg climbs to 3 or 4, and Zabbix server keeps rotating the log file because it seems to produce a megabyte of log every few seconds!

Once I inactivate the new host in the web interface, everything comes back to normal.

I think there is a bug in the handling of missing Zabbix agents, the server should not go crazy like this if a single agent is missing.

Regards

[Alexei]

We are aware of this issue. It is related to some issues in a new networking code we introduced in 1.3.5. Thanks for reporting.