[v1.29] ephemeral-storage-quotas: repromote to beta

[pacoxu]

#1029: Quotas for Ephemeral Storage

• One-line PR description: re-promote LocalStorageCapacityIsolationFSQuotaMonitoring feature to beta

• Issue link: Quotas for Ephemeral Storage #1029

• Other comments: Quotas for Ephemeral Storage #1029 (comment)

This is discussed in recent sig-node meetings.

• The last promote PR: promote ephemeral-storage-quotas to beta in 1.25 #2697, and I think this feature is almost meets the beta bar.
• https://testgrid.k8s.io/sig-node-kubelet#kubelet-gce-e2e-fsquota-ubuntu : test grid was added.
• • the blocker is fixed in kubelet: Fix fs quota monitoring on volumes kubernetes#115314.

Pros:

• without this feature, du in some scenarios will hang or take too much time.
• metrics were added add VolumeStatCalDuration metrics for fsquato monitoring benchmark kubernetes#107201
• add warning log if volume calculation took too long than 1 second kubernetes#107490
• the blocking issue mentioned is fixed as above.

Cons:

• if quota id is changed in the kernel layer, is there any potential security problem?

I prefer to promote this if there is no new issue raising about this feature. The last promotion helped us find the bug kubernetes/kubernetes#112624 and kubernetes/kubernetes#115314 that we fixed recently.

• If there are still other concerns about promoting the feature, we may postpone deciding whether to promote it in v1.28.

[pacoxu]

/assign @rphillips @mrunalp

[bart0sh]

/lgtm

[pacoxu]

The CI can pass with kubernetes/kubernetes#115314.

• I tested it in [pending UserNamespace beta]promote LocalStorageCapacityIsolationFSQuotaMonitoring to beta kubernetes#112626

If the blocking issue is fixed, I still prefer to promote it to beta in v1.27 or the future release.

[pacoxu]

Ping @rphillips @mrunalp

/assign @SergeyKanzhelev

[pacoxu]

Any new thoughts on this @rphillips

[bart0sh]

/lgtm
/assign @derekwaynecarr @dchen1107

[SergeyKanzhelev]

Gradutaiton criteria for beta:

• User feedback
• Benchmarks to determine latency and overhead of using quotas relative to existing monitoring solution
• Cleanup

Is there anything you can share w.r.t. the above? Perhaps add it to this readme file.

Also test plan list one e2e test. Do we need to have a test that validates file systems that do not has this capability? Is it something we need to test?

[pacoxu]

The test was added at kubernetes/test-infra#28616.

Some general test of the performance benchmark is like below. And the metric was added kubernetes/kubernetes#107201.

• for 2 Gi volume, without this FG, it takes 1-2 minutes. with the FG enabled, it takes 1-10ms.

Cleanup is vague to me.

User feedback part: @rphillips do you have some? We enabled this in some test env and only few customers give it a try in some very old version kubelet.(I suppose we should not add the detailed feedback here.)

[pacoxu]

Also test plan list one e2e test. Do we need to have a test that validates file systems that do not has this capability? Is it something we need to test?

Do you mean that we should do some fall back test when this is not supported? The fall back behavior is everywhere, I think.

[pacoxu]

ACK @SergeyKanzhelev do you have some further comments?

[SergeyKanzhelev]

/lgtm

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: mrunalp, pacoxu, SergeyKanzhelev

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• keps/sig-node/OWNERS [mrunalp]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment