Victoriametrics: Provide component/app label in cluster version

[oOHenry]

Is your feature request related to a problem? Please describe.
I use the victoria metrics cluster grafana dashboard. On the dashboard is an uptime panel but the uptime is only listed by the instance label. It would be nice if the table would show the uptime by app vmstore, vmselect and vminsert, but this isn't possible because the vm_app_start_timestamp has no app label.

Describe the solution you'd like
Metrics which are the same in vminsert, vmstorage and vminsert e.q. vm_app_start_timestamp should have an app or component label. That a user can see which cluster component has delivered this metric.

is:

vm_app_start_timestamp 1573829942
vm_app_uptime_seconds 242680

should:

vm_app_start_timestamp{component="vmstore"} 1573829942
vm_app_uptime_seconds{component="vmstore"} 242680

[valyala]

This can be achieved by grouping by job label, which is set by Prometheus according to scrape configs. @hagen1778 , could you look into this?

[oOHenry]

@valyala OK but this requires to have a separate scrape job for each component.

[valyala]

It is recommended to have separate scrape jobs in Prometheus for vminsert, vmselect and vmstorage components, since their metric naming can clash as with vm_app_uptime_seconds mentioned above.

[hagen1778]

Hi @oOHenry !
I totally agree with @valyala - it is recommended to have separate scrape jobs. Pls see example of Prometheus configuration in our docker-compose env. This also helps to separate process metrics for each application:

So with separate jobs Uptime panel should work as expected:

[oOHenry]

Ok, thanks for clarifying. I think we can close this issue.