Agent status on the Zabbix-server doesn't work/differs from other agents

[Casn]

Hi,

i am building a monitoring-server with your great software. So far I have 77 Hosts monitored with many different checks and templates.
Zabbix-version is 1.6.5 on all systems (server and agents)

All clients report their agent status (Item Type "ZABBIX agent", Key "status", Update interval "30") as expected. Only the agent on the Zabbix-server itself doesn't change its value (no data) even if I stop the agent or start it again later.

I'm aware of the fact, that the value is only updated, if there is a change between available/not available. See Alexei's Post:

http://www.zabbix.com/forum/showpost.php?p=22192&postcount=10

The problem is, that not only this item and the associated trigger does not work, but also the other triggers don't recognize that the agent is down and still show the last values the server received (they stay ok). On working servers they have this "default dependency" that they will show up in grey in the overview etc.

What i have tried to fix this:
- deactivate/activate the Item for the Server -> no change
- change update interval
- restarting zabbix_server process(es)
- zabbix_agentd.conf:
- Change debuglevel to 4 -> no useful information
- Set Server=127.0.0.1 and 192.168.200.210 (eth0-interface)
- and many absurd things ;-)

If have no clue, why this doesn't work on just this host or if it's even a bug. Maybe someone has an idea!?

Regards

Carsten

[richlv]

what kind of item types are you using for that host ? .status is only updated for normal/passive agent items.

[Casn]

As mentioned, the "status"-Item has type "ZABBIX agent". I also tested "ZABBIX agent (active)" -> same result: All other agents report their status and the other (dependent) triggers get greyed out, but only the agent on the Zabbix-Server does not report that it's down and all other triggers with the old data stay OK.

[richlv]

no, what item types do _other_ items have - do you have at least one working zabbix agent type item or snmp item ?

[Casn]

Ok, sorry for misunderstanding.

And yes, there are several working items (ZABBIX agent, Simple check, ZABBIX internal). I have attached a screenshots of most of the items and the latest data of the application "General" (had to resize it). And there doesn't appear the "Agent status"-item, because of no data.

And btw: Is the "Availability"-column on the Hosts-page depending on at least one working zabbix agent or snmp item? That would explain the "Unknown" entries on hosts which are only pinged...

[Casn]

Here are the results from some additional tests:

• an item with key "agent.ping" and type "ZABBIX agent" stays always 1 (Ok, as said in documentation) and neither item nor trigger gets "unsupported" or false or something like that
• the simple check "tcp,10050" recognizes that the agent is not running an the appropriate trigger fires. But the "status" still has no data which results in the already mentioned problems/misbehavior.

Only workaround I plan now is to take the "tcp,10050" item with trigger and make all other triggers for this server explicit dependent on that one.
But it seems still like a bug for me...

[richlv]

agent.ping only gets data when agent is active, thus to find out agent inactivity based on this item you have to use nodata() function

[Casn]

Yeah I read about it... simply forgot it

Workaround is ok even though not handsome in the overview.

And btw : Triggers massupdate can't be used to remove a dependency (always says "No dependencies defined")...

[bek99]

I'm seeing this too. The problem is that I'm not seeing this across the board on all the machines. I have a machine fully offline right now, but for whatever reason and the zabbix server thinks it's fine and available.

http://www.zabbix.com/forum/showthread.php?t=12982

I made some references.. Currently major difference I have is I have main zabbix server 1.6.5 (but did try 1.6.4 and same thing) on i386 and all the agents are x64. Everything fires off fine tests wise and such, but if the agents disappear on the machines on same physical subnet, no reports, but if they disappear across the vpn, it picks it up.

Now, on a second instance I have running that's x86_64, it works fine across the board (using 1.6.4).

All are ubuntu hardy server. so am at a loss, I can probably follow your lead for the tcp check, but that isn't a nice way to roll. :/

-b