Capture metrics for external volume resizing

[gnufied]

No description provided.

[fejta-bot]

Issues go stale after 90d of inactivity.
Mark the issue as fresh with /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.

If this issue is safe to close now please do so with /close.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/lifecycle stale

[bertinatto]

/remove-lifecycle stale.

[bertinatto]

What metrics should the external-resizer emit?

If I read correctly, in-tree CSI resizing already emits metrics for duration, errors, statuses etc.

I suppose the duration there isn't exactly accurate (because the resizing is performed externally), is that correct? If so, would recording the duration of the operation enough here?

[msau42]

@yuxiangqian is planning to add a metrics library to csi-lib-utils that all the sidecars can use

[msau42]

The main benefit of sidecar metrics is to get per operation latency and error metrics which cannot be done in the intree controller for CSI plugins

[gnufied]

@bertinatto we do not emit resizing metrics for CSI drivers from in-tree code. Where did you see the metric being emitted? The in-tree controller simply hands over the control to the external controller.

@msau42 Will the library capture just operation metrics or total latecy metrics?

[fejta-bot]

Stale issues rot after 30d of inactivity.
Mark the issue as fresh with /remove-lifecycle rotten.
Rotten issues close after an additional 30d of inactivity.

If this issue is safe to close now please do so with /close.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/lifecycle rotten

[msau42]

/lifecycle frozen

[yuxiangqian]

@gnufied @msau42 I have not got a chance to work on the lib yet. I will come back to that after the snapshot to beta project is done. But the plan is to have it in csi-lib-utils. My very initial thought is to have common utilities, like metric registration with prometheus, timestamp cache struct etc. defined for sharing. The real reporting would still happen likely in sidecar controllers. Might worth a KEP?