How be sure zabbix_agentd is running ?

[Kai-Kai]

Hello,

I use the item "status" to be sure the zabbix_agentd is running. And the trigger is status is different from 0.
But sometimes, I receive an alert whereas the agent is not really down...

So, I would like to better the trigger, to avoid having a useless notification.

The current trigger is :
{Machine:status. last(0)}#0

I've tried :
{Machine:status. last(180)}#0

My goal si to activate the trigger only if the status is different from 0 for more than 180 seconds...

But the trigger continue to alert immediatly when the agent is down.

I've also try agent.ping, but the agent.ping doesn't receive anything if the agent is down... and if I try to count number of "1" during the last 180 hours, or if I try to use .nodata to detect if we receive values or not... it's seems not to work properly.


So... my questions are :
* how do you monitore the agent is running ?
* is there a way to use status, considering the time for example (alert if the status is different from 0 during more than X seconds) ?


Thanks in advance for your help.

[Kai-Kai]

On the page 130, in the pdf documentation, I can read :

Example 7 Server is unreachable
{zabbix.zabbix.com:status.last(0)}=2
Note: The ‘status’ is a special parameter which is calculated if and only if
corresponding host has at least one parameter for monitoring. See description of
‘status’ for more details.

But... where's is this "description of ‘status’" mentionned here ? I can't find it... for example to know if last(180) is supported, if another thing is supported ?...

Thx

[Calimero]

We use a .nodata(X) trigger on agent.ping to check whether zabbix_agent is alive.

We're changing our templates these days so we may actually add an extra "icmpping" check/trigger (the agent.ping trigger would depend on it to prevent dual notifications).

status is a bit of a "virtual" item whose value is maintained by zabbix_server depending whether remote checks (snmp, ping, agent) where successful or not.

status isn't refreshed every X seconds. You only have the last value (last status change) and previous value in the items table. There's no history. So .nodata() won't work as expected.

I personally avoid using status in my triggers and rely on specific triggers (.nodata on an SNMP item for network devices, .nodata on agent.ping for zabbix_agentd-enabled servers...).

[Kai-Kai]

I think you're right. I've replaced my status item by a agent.ping one.
I've also read in the documentation that seconds number in last is ignored... ^^
I've made a trigger on :
****.agent.ping.nodata(180)#1

I think it will avoid sending notifications when it should not.

Thanks for your explaination.

[Calimero]

I think you're right. I've replaced my status item by a agent.ping one.
I've made a trigger on :
****.agent.ping.nodata(180)#1

Mhh... =1 (instead of #1)

[Kai-Kai]

Yes sorry.
In fact, my trigger is :
{****:agent.ping. nodata( 180 ) }#0
to be precise.

I've answered on the forum without verifying the exacte value in my zabbix. Sorry.

[Kai-Kai]

I've finally modified it :
{****:agent.ping. nodata(500) }=1
I think it's better...

Thank you for your discussion. The same problem happened to me. It is very helpful to me.

Really happy to have helped you.