Internal item type

[[email protected]]

This may be a dumb question/observation but maybe someone can clarify. Internal item types (in 3.4.x), particularly for monitoring Zabbix internal processes (history syncer, configuration syncer, etc.) will only run on the system that Zabbix Server or Zabbix Proxy is running, and in the case of a Zabbix proxy, it has to monitor itself to get those values.

Is there no way to get those values remotely?

The documentation isn't totally clear on this

Internal checks are calculated:

• on Zabbix server - if the host is monitored by server
• on Zabbix proxy - if the host is monitored by proxy

Reason I noticed this is my Zabbix Server, was configured to monitor itself, and all of it's Proxies. However all of the internal items for the proxies, matched the Server. I had to set them to monitor them selves (setting Monitored by Proxy box on the host configuration for the Proxy, to itself) to get this to work correctly.
Also I setup another Zabbix Server, to monitor my production Zabbix Server that I believe is suffering from a bug fixed in 3.4.8, but it was only pulling it's own values for internals. How would you monitor Zabbix Server internals, from another Zabbix Server?

[kloczek]

This may be a dumb question/observation but maybe someone can clarify. Internal item types (in 3.4.x), particularly for monitoring Zabbix internal processes (history syncer, configuration syncer, etc.) will only run on the system that Zabbix Server or Zabbix Proxy is running, and in the case of a Zabbix proxy, it has to monitor itself to get those values.

Is there no way to get those values remotely?

The documentation isn't totally clear on this

Internal checks are calculated:

• on Zabbix server - if the host is monitored by server
• on Zabbix proxy - if the host is monitored by proxy

To be more precise. those metrics are sampled by zabbix server and proxy processes.

There is 3rd type of internal metrics as well which is related to the state of the monitored zabbix host .. which is hard to classify because it is sampled by the server processes but it presens exact host state.
Such metric is "zabbix[host,,items_unsupported]" and it gives number of items on the exact host where this metric is added which holds counter with items in unsupported state. Such metric is extremely useful when for some reason sampling some metrics on the host starts failing (backend method is failing or even someone made typo on modify/maintaining template or if host collects data from SNMP device it may start showing some non-zero value on monitored device with SNMP agent because for example some changes done in new version of the device firmware) [1].
You can find this metric in use in my template https://github.com/kloczek/zabbix-te...Zabbix%20Agent.
This metric is present as well on templetes available here https://github.com/kloczek/zabbix-te...SNMP%20Devices

(TBH I know about this metric because few years ago I've proposed patch which added this internal metric :P)

[1] Recently I figured that every SNMP agent must support SNMPv2 MIB and probably it would be better to move this metric to SNMPv2 MIB based template https://github.com/kloczek/zabbix-te...MIB/SNMPv2-MIB

Reason I noticed this is my Zabbix Server, was configured to monitor itself, and all of it's Proxies. However all of the internal items for the proxies, matched the Server. I had to set them to monitor them selves (setting Monitored by Proxy box on the host configuration for the Proxy, to itself) to get this to work correctly.
Also I setup another Zabbix Server, to monitor my production Zabbix Server that I believe is suffering from a bug fixed in 3.4.8, but it was only pulling it's own values for internals. How would you monitor Zabbix Server internals, from another Zabbix Server?

As you already spotted it is not possible to have two zabbix servers in whole zabbix monitoring configuration and obviously it is not possible to monitor two zabbix servers internal metrics over one set of monitoring configuration metrics using internal zabbix item type.
If other zabbix server process will be down it will be nothing which will be collecting this server new internal metrics data so such cross monitoring IMO has no sense. So that is from point of HA.
From point of view of the state of the whole stack as long as server is up ad running new data about stack state will be collected, stored and presented/available over web frontend or API.
Even if server process will be down and DB backend will be up and running you will be able access to those internal metrics historic data without fiddling with DB backend SQL queries. Cloning those data IMO (again) does not make to much sense.
Proxy internal metrics data are sampled by proxy process so even if it is disconnection between srv<>prx still new data from those metrics will be collected. When prx<>srv connectivity will be restored all data will be pushed to the server.