We're getting Trigger Alerts when we should not

[just2blue4u]

Hi there!

We keep getting alerts (and upalerts) for systems that have been switched off since Friday.

They say "CPU Load too high" or "Zabbix agent unreachable for over 2 minutes", or "Zabbix agend reachable again" (as said- that system is still switched off).

System Load of our zabbix server is ok, the mysql-db seems happy, too.
I restarted the zabbix server program but that didn't help!
When i see the history of actions, all Alerts are listed as "not sent" with 3 retries left. We do send them via email and windows popup (selfmade shellscript using samba).

We do get alerts, so mail and script do work good, but why is zabbix sending us this crap? A switched off system can't switch between reachable and unreachable all the time; it should be unreachable all the time!

What can i do to get rid of that alerts?

[richlv]

...
When i see the history of actions, all Alerts are listed as "not sent" with 3 retries left. We do send them via email and windows popup (selfmade shellscript using samba).

could it be that your method of sending alerts does not return correct exit code, so zabbix assumes the message was not sent successfully ?

[just2blue4u]

since when does Zabbix try to re-send unsent alerts ?

We send via email and with this script:

#!/bin/bash
#
# Send a zabbix message via winpopup to a given host
# Use as an alertscript within zabbix
# Syntax: winpopup $receiptant $subject $message
# Note: zabbix passes $1=receiptiant, $2=subject, $3=message to any script
echo -e "=============================================================\\n$2\\n=============================================================\\n\\n$3"  | smbclient -M $1  -U="zabbix imv neu"

Today this problems occured first. We didn't do anything with our server (no update of zabbix server bins)

[richlv]

since when does Zabbix try to re-send unsent alerts ?

i haven't investigated this, i don't remember having failed sending attempts...

We send via email and with this script:
...

try to run the script as zabbix user with the exact parameters that are used in problematic action.

[just2blue4u]

that problem stopped and didn't come again.
Seems like it was the server being confused after a server and db restart.

Is this a bug or was it just bad luck?

[richlv]

do these triggers have 'nodata' in their expressions ?

[just2blue4u]

yes:
{<host>roc.num[zabbix_agentd].nodata(120)}

[richlv]

which zabbix version ?
there was a bug involving nodata triggers and zabbix server restarts, which is supposed to be fixed in 1.4.5 (though i still got the problem once after upgrading to 1.4.5 - kinda unwilling to reproduce, having several hundred warning emails sent around was bad once...)

[just2blue4u]

Server Version is 1.4.4 (just updated my sig! ). Clients are different...

[richlv]

would be useful to check out whether you see this behaviour with 1.4.5 as well.

[just2blue4u]

i'm writing an insttallscript for clientinstallation on rpm-based dell machines.
Planned to do the Server update after that!