What exactly ARE pollers??

[ebonweaver]

I'm trying to figure out why the "unreachable poller" process is so busy on the server. That leads me to trying to figure out what the pollers are to begin with. The documentation indicates these are only needed if you have IPMI or Java poller, but that's not true. If you set StartPollersUnreachable=0 the service no longer starts at all, so it's clearly required. I have 15 going and they stay 25% busy all the time. But what are they, what are they doing, and why?? The only mention I find relates to Discovery, which we're not using (no rules at all, anywhere). We used some for a while to scan subnets for hits, but then turned them off, so unless something is wicked stuck in the server, that's not it. The only other thing I found is related to missing hosts. There is one host based on the log that the server is trying to send active checks to. That host does not exist. Like, at all, it's not anywhere in the host configurations in the server. So, this clearly IS an instance of something being stuck (it used to exist, 3 years ago...) that failed to clear out for some reason. But that one item can't be throttling these unreachable poller processes, so what's up with these things?

[mtjzb]

Pollers are processes that the main zabbix_server process starts up to do things. Unreachable pollers are dedicated processes that check your "unreachable" hosts periodically (set in your config file), to see if they've become reachable. If you have 15 unreachable pollers, and they're 25% active, you likely have a lot of hosts configured in zabbix that are showing as down for some reason.

[kernbug]

I'm trying to figure out why the "unreachable poller" process is so busy on the server. That leads me to trying to figure out what the pollers are to begin with. The documentation indicates these are only needed if you have IPMI or Java poller, but that's not true. If you set StartPollersUnreachable=0 the service no longer starts at all, so it's clearly required. I have 15 going and they stay 25% busy all the time. But what are they, what are they doing, and why?? The only mention I find relates to Discovery, which we're not using (no rules at all, anywhere). We used some for a while to scan subnets for hits, but then turned them off, so unless something is wicked stuck in the server, that's not it. The only other thing I found is related to missing hosts. There is one host based on the log that the server is trying to send active checks to. That host does not exist. Like, at all, it's not anywhere in the host configurations in the server. So, this clearly IS an instance of something being stuck (it used to exist, 3 years ago...) that failed to clear out for some reason. But that one item can't be throttling these unreachable poller processes, so what's up with these things?

Hi,

Could you provide the following graphs?

• Zabbix server performance
• Zabbix internal process busy
• Zabbix data gathering process busy
• Zabbix cache usage
• Number of unreachable hosts/unsupported items

[ebonweaver]

Ok so first thing then is, the documentation is inaccurate, unreachable pollers throttle constantly for ANY unreachable host of ANY type. I proved this by disabling a couple AWOL printers. It appears that devices offline for service are a serious impact on this, and there needs to be a performance setting somewhere to mitigate this. One down printer should not throttle the server at default settings. I don't think I should need 50 pollers running in an enterprise because they each throttle one device that may be offline for any number of fine reasons.
In regards to the graphs, I'm immediately curious about the mention of one that would show unreachable hosts and unsupported items. That sounds very useful, but it does not exist from what I see. Given the other graphs also did not exist until I found a template to add to the system, I'm guessing it's a similar thing. Better yet, some way to list out such items for investigation and clean up. Can you point me to such resources?

[kernbug]

Hello

Information about unreachable/unsupported:

https://www.zabbix.com/documentation/3.4/manual/appendix/items/unreachability

Also check this threads:

https://www.zabbix.com/forum/zabbix-troubleshooting-and-problems/50716-zabbix-unreachable-poller-processes-more-than-90-busy

https://www.zabbix.com/forum/zabbix-troubleshooting-and-problems/50469-zabbix-poller-processes-more-than-75-busy?t=58102

[Pellegrin59]

You have Additional Pollers installed. Additional Pollers allow you to increase the monitoring capacity of your Orion installation by enabling multiple monitoring engines that work in parallel across your network firstbankcard.