Trying to monitor a system the occassionally half crashes.

[scalft]

Hello,

I have a Linux server that will occasionally kernel panic, but not take the system completely down. The server will still ping, but will not accept TCP connections (ssh, zabbix agent, etc.). I am trying to design a way to effectively monitor for this situation. I would also like it to work from a template so I don't have to go through about 30 servers to configure each individually.

I have attempted

tcp,10050

as a simple check on a test host.

When I stop zabbix_agent on the host, the value sometimes changes to zero, but sometimes doesn't. Checking zabbix_server.log I see the following:

  4961:20080515:092422 Host [hostname]: another network error, wait for 15 seconds
  4961:20080515:092437 Get value from agent failed. Error: Cannot connect to [a.b.c.d:10050] [Connection refused]
  4961:20080515:092437 Host [hostname] will be checked after 60 seconds
  4961:20080515:092547 Timeout while answering request
  4961:20080515:092547 Get value from agent failed. Error: Cannot connect to [a.b.c.d:10050] [Interrupted system call]

Meaning the host is not being checked during that interval... Therefore no update of my simple check item, and subsequently no trigger.

Thoughts?

Thad

[nelsonab]

you could trigger on nodata key[item].nodata(time)