RFE: Replace ServiceMonitor for PodMonitor

[camilamacedo86]

What do you want to happen?

It was suggested by @erikgb :

Less resources is better IMO, so I suggest adding an additional port to the manager service or use PodMonitor instead of ServiceMonitor.

Reason: We need to keep the args in manager and in the patch otherwise when the patch is made they are lost. More info; #3934

The goal of this task is to analyse this possibility and see if we can came with a better solution.

Extra Labels

No response

[joelanford]

One thing came to mind after we chatter about this. Would PodMonitor be a problem for multi-replica controllers that use leader election? prometheus-operator/prometheus-operator#3119 (comment)

Probably need to try it out, but I wonder if PodMonitor and ServiceMonitor would result in different sets of scrape targets in our situation. Do controller-runtime non-leader replicas report liveness? And if not, I assume they are excluded from the scrape targets of ServiceMonitor. Is the same true for PodMonitor?

[camilamacedo86]

In the latest changes, following the @joelanford great ideas and suggestions I think we could either sorted the motivation here now we can add new args without the need to duplicate them in 2 files so I think we can close this one. However, if anyone think that is valid persue this idea please feel free to re-open and add your thoughts with why we would need to do so.

Thank you for all collaboration !!!