too many false alarms about server unreachable

[harri]

Hi folks,

I get too many false alarms about "Server somehost is unreachable". Of course its not always the same host. The network is fine. After a minute the server is back, but by then Zabbix already notified everybody by EMail.

How comes? Is there some way to tell Zabbix to try a little bit harder before sending out an alarm first and a "nevermind" 30 seconds later?

Sorry for asking, but the Zabbix book doesn't tell. Any helpful comment would be highly appreciated.


harri

[wizard-ict]

You may find this thread useful

http://www.zabbix.com/forum/showthread.php?t=25055

[harri]

There was another thread mentioning something like {mytemplate:status.count(120,2,"eq")}>1, but that doesn't really solve the problem. Its just a delay and a second try before sending out the alarm. I don't like the delay in case of a real problem.

I wonder why Zabbix thinks the server is unreachable when it is not?

[wizard-ict]

In my (limited) experience it is just a timeout waiting for data when the host is busy. You can increase the timeout period in the zabbix_server.conf file which might solve your problem

[Jaeius]

I suggest you adjust/change the Expression you are using for "Server {HOSTNAME} is unreachable" Trigger.


You can use expressions like:

{HOSTNAME:agent.ping.nodata(300)}=1 --> agent unreachable via ping for 5 mins

{HOSTNAME:icmpping.count(360,0)}>5 --> host unreachable 5 times for the past 6 mins