Host group isssues appearing on the wrong hosts

[duncan9562]

I have a few proxies which are all in a host group, called Zabbix Proxies.

For arguments sake the proxies are called:-
proxy1
proxy2
proxy3

proxy1 has a fair few hosts assigned to it, and occasionally gets the "Zabbix poller process more than 75% busy" which is fine.

The issue I have, is that when that trigger is triggered on proxy1, it also shows up on proxy2 and proxy3 as "Zabbix poller process more than 75% busy". This also shows up under the Zabbix Server host group, but all relates to proxy1 being overloaded. Both the other proxies and the server are under no load what so ever.

Currently proxy2 and proxy3 have no checks assigned, and looking at the Zabbix process list I can see they are processing absolutely nothing, so this cannot be true.

Has anyone seen this before / have any ideas as to what I may have done?

Thanks in advance.

[allexpetrov]

Hi duncan9562,

can you post the trigger configuration ?

Regards,
Alex!

[duncan9562]

Hi Alex,

Thanks very much for your swift response!

Sure thing, please see below:-

Name: Zabbix poller processes more than 75% busy
Problem expression: {zbx-rd-prx1:zabbix[process,poller,avg,busy].avg(10m)}>75
Recovery expression: {zbx-rd-prx1:zabbix[process,poller,avg,busy].avg(10m)}<65

To give you a bit more info that might be useful, I have a template set up called "zabbixproxy". I have linked the "Template App Zabbix Proxy" to the template I created (zabbixproxy). I then have the zabbixproxy template assigned to 4 proxies within my environment.

Hopefully this helps, if you need any more info please do let me know.

Duncan.

[vendrusculo]

Zabbix performance items are "Zabbix Internal" it means that the server/proxy doesn't ask these values to the host...

To fix it just change the host/proxy to be monitored by it self

[duncan9562]

Zabbix performance items are "Zabbix Internal" it means that the server/proxy doesn't ask these values to the host...

To fix it just change the host/proxy to be monitored by it self

Thanks very much for your response vendrusculo.

I do actually have the proxies monitored by themselves. I.e, proxy1 is monitored by proxy1 and so on so forth.

Apologies if this wasn't what you meant....

[vendrusculo]

it was exactly what I tried to say... but if you already it done I have no clue what can be... Have you already compared the graphs/data between the proxies, are they different?

[Semiadmin]

Hi duncan9562!
Your recovery expression is wrong. If you want to fire a trigger on 75% and clear it on 65%, recovery expression must be
{zbx-rd-prx1:zabbix[process,poller,avg,busy].avg(10m)}>65

[duncan9562]

Hi all,

Thanks for your replies.

Your recovery expression is wrong. If you want to fire a trigger on 75% and clear it on 65%, recovery expression must be
{zbx-rd-prx1:zabbix[process,poller,avg,busy].avg(10m)}>65

The trigger works absolutely fine, the fact is that, I have proxy1 and proxy2. proxy1 has a fair number of checks / host assigned to it, and proxy2 has absolutely none. When proxy1 gets busy, the same trigger is triggered on both proxies, even though proxy2 is doing nothing!

it was exactly what I tried to say... but if you already it done I have no clue what can be... Have you already compared the graphs/data between the proxies, are they different?

The graphs are different, however they do show similar load spikes etc in around the same places which I find weird. The fact that proxy1, proxy2 and server1 show the same trigger when prox1 is under load is strange. I would expect that, if proxy1 was overloaded it would only trigger the triggers assigned to prox1.

I must have something configured wrong but I really am struggling to work it out!

Thanks again guys.

[allexpetrov]

Hi,
I assume that you are using the default templates, yes? Could you please show how the problem lookslike(screenshot) and name you version of server, proxy and agent.

Regards,
Alex!

[duncan9562]

Hi Alex,

I am indeed using the default template. This template is applied to all Proxies. The Zabbix Server template is applied to the server as well.

Sure, please see the screenshot below:-


So the proxy in question is zbx-rd-prx1, which has all the checks assigned to it, and should be the only one that triggers the "Zabbix Poller process more tha 75% busy" however, this seems to trigger for all three, zbx-rd-prx1, zbx-rd-prx2 and zbx-rd-srv1.

The versions for all are '1:3.2.7-1+jessie'.

Thanks again.

[allexpetrov]

Maybe a little bit stupid question, but it looks like your proxies are installed on the same OS machine as the server. Is it possible?

Regards,
Alex!

[duncan9562]

Hi Alex,

The proxies and servers are running on completely separate nodes and have completely separate hostnames.

Thanks again for your continued efforts.