Host availibility Trigger flaps even though host is up

[jcesario]

Im using agent.ping for host availability.

My item is

agent.ping @ a 10 second interval

my trigger is

agent.ping.nodata(60)

--

The trigger keeps flapping true. Im not understanding why but when trying to graph the agent.ping theres plenty of holes in the graph line.

How do I go about troubleshooting this?

[neurovish]

I got weird results when I used agent.ping as well. I'm using icmpping now and it seems to work well. I have my unavailable trigger set to "icmpping.count(#4,#1)}=4" which will mark a host as unavailable if it host doesn't respond for 4 consecutive checks.

[jcesario]

I got weird results when I used agent.ping as well. I'm using icmpping now and it seems to work well. I have my unavailable trigger set to "icmpping.count(#4,#1)}=4" which will mark a host as unavailable if it host doesn't respond for 4 consecutive checks.

This returns an error as an invalid trigger expression.

I cant seem to find in the manual where count accepts function arguments....

Can you point me to some documentation?

[neurovish]

pg. 125 in the Zabbix 1.6 manual for count, and pg. 114 for icmpping. I'm also running version 1.6.4.

[jcesario]

Thanks. Found it. I was looking at the wrong "count"

[jcesario]

Is there a bug in agent.ping or something?

The weird thing is theres no holes in all the other graphs for the host. Just for agent.ping....

[neurovish]

I don't know of any specific bug, icmpping just worked better and reported what I was interested in more (server availability, not necessarily agent availability) anyways. I'm not sure what is causing the holes.

[Alt]

Am running 1.6.1 and use agent.ping for host availability. Am experiencing Host availability flaps on some servers. Other servers are fine. Does anyone have any idea on what might be causing this?

[Sparky]

Not sure if this directly relates to the issues here, but I found that when using agent ping, even if 1 packet of the test ping fails because of network congestion, time out etc etc, the entire check will fail.

The equipment I am checking, by design, will not return all ping packets, so instead I use the icmppingloss check. Works like a charm for me. Only if 100% of packets are lost will it raise the trigger.