You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Items marked with (R) are required *prior to targeting to a milestone / release*.
37
37
38
38
-[x] (R) Enhancement issue in release milestone, which links to KEP dir in [kubernetes/enhancements] (not the initial KEP PR)
39
-
-[] (R) KEP approvers have approved the KEP status as `implementable`
40
-
-[] (R) Design details are appropriately documented
39
+
-[x] (R) KEP approvers have approved the KEP status as `implementable`
40
+
-[x] (R) Design details are appropriately documented
41
41
-[ ] (R) Test plan is in place, giving consideration to SIG Architecture and SIG Testing input (including test refactors)
42
42
-[ ] e2e Tests for all Beta API Operations (endpoints)
43
43
-[ ] (R) Ensure GA e2e tests meet requirements for [Conformance Tests](https://github.com/kubernetes/community/blob/master/contributors/devel/sig-architecture/conformance-tests.md)
44
44
-[ ] (R) Minimum Two Week Window for GA e2e tests to prove flake free
45
-
-[] (R) Graduation criteria is in place
45
+
-[x] (R) Graduation criteria is in place
46
46
-[ ] (R) [all GA Endpoints](https://github.com/kubernetes/community/pull/1806) must be hit by [Conformance Tests](https://github.com/kubernetes/community/blob/master/contributors/devel/sig-architecture/conformance-tests.md)
47
-
-[] (R) Production readiness review completed
47
+
-[x] (R) Production readiness review completed
48
48
-[ ] (R) Production readiness review approved
49
-
-[] "Implementation History" section is up-to-date for milestone
49
+
-[x] "Implementation History" section is up-to-date for milestone
50
50
-[ ] User-facing documentation has been created in [kubernetes/website], for publication to [kubernetes.io]
51
51
-[ ] Supporting documentation—e.g., additional design documents, links to mailing list discussions/SIG meetings, relevant PRs/issues, release notes
52
52
@@ -107,7 +107,7 @@ to allow querying specific pods for their allocated resources.
107
107
returns the list of PodResources for *all* pods across *all* namespaces in the
108
108
cluster). That is, it allows one to specify a specific pod and namespace to
109
109
retrieve PodResources from, rather than having to query all of them all at
110
-
once.
110
+
once.`Get()` returns error if the pod is known to the kubelet, but is terminated.
111
111
112
112
The full PodResources API (including our proposed extensions) can be seen below:
113
113
@@ -274,8 +274,9 @@ These cases will be added in the existing e2e tests:
274
274
275
275
#### Beta
276
276
277
-
-[ ] Gather feedback from consumers of the DRA feature.
278
-
-[ ] No major bugs reported in the previous cycle.
277
+
-[x] Gather feedback from consumers of the DRA feature.
278
+
- Integration with the NVIDIA DCGM exporter (https://github.com/NVIDIA/dcgm-exporter/pull/501) to gather per pod Dynamic Resources managed by [k8s-dra-driver-gpu](https://github.com/NVIDIA/k8s-dra-driver-gpu).
279
+
-[x] No major bugs reported in the previous cycle.
279
280
280
281
#### GA
281
282
@@ -333,7 +334,7 @@ The API becomes available again. The API is stateless, so no recovery is needed,
333
334
334
335
###### Are there any tests for feature enablement/disablement?
335
336
336
-
e2e test will demonstrate that when the feature gate is disabled, the API returns the appropriate error code.
337
+
e2e test will demonstrate that when the feature gate is disabled, the API returns the appropriate error code. (https://github.com/kubernetes/kubernetes/pull/116846)
337
338
338
339
### Rollout, Upgrade and Rollback Planning
339
340
@@ -347,7 +348,12 @@ Kubelet may fail to start. The new API may report inconsistent data, or may caus
347
348
348
349
###### Were upgrade and rollback tested? Was the upgrade->downgrade->upgrade path tested?
349
350
350
-
Not Applicable.
351
+
Not Applicable. Because this change:
352
+
353
+
- Is read-only in the kubelet’s in-memory state.
354
+
- Is behind a feature gate, so turning it off simply disables the new endpoints without affecting any existing behavior.
355
+
356
+
In practice, restart the kubelet with the gate disabled (rollback) or re-enabled (upgrade), and the API behavior reverts or returns without loss of data or consistency. Therefore we don’t need a special upgrade/downgrade test matrix for this KEP.
351
357
352
358
###### Is the rollout accompanied by any deprecations and/or removals of features, APIs, fields of API types, flags, etc.?
353
359
@@ -372,7 +378,9 @@ Call the PodResources API and see the result.
372
378
373
379
###### What are the reasonable SLOs (Service Level Objectives) for the enhancement?
374
380
375
-
N/A.
381
+
100% in normal operation. The proposed API exposes in read only mode kubelet internal data, critical for functioning of the kubelet.
382
+
This data has to be available 100% of the time for the proper functioning of the kubelet, thus is expected to be available 100% of time.
383
+
The only possible error source is the API calls being throttled by the rate-limiting introduced with the GA graduation of the parent KEP 606.
376
384
377
385
###### What are the SLIs (Service Level Indicators) an operator can use to determine the health of the service?
378
386
@@ -408,36 +416,48 @@ No.
408
416
409
417
###### Will enabling / using this feature result in increasing size or count of the existing API objects?
410
418
411
-
No.
419
+
No. Enabling this feature does not change the number of API objects returned. But it may increase the size of each object whenever there are Dynamic Resources to report where each ContainerResources now has an extra dynamic_resources field.
412
420
413
421
###### Will enabling / using this feature result in increasing time taken by any operations covered by existing SLIs/SLOs?
414
422
415
423
No. Feature is out of existing any paths in kubelet.
416
424
417
425
###### Will enabling / using this feature result in non-negligible increase of resource usage (CPU, RAM, disk, IO, ...) in any components?
426
+
Negligible amount of CPU and memory. Because the API is purely read-only and piggy-backs on the kubelet’s existing cache and checkpointing machinery, exposing Dynamic Resources incurs only similar minimal serialization and storage as CPUManager and DeviceManager—so any extra CPU, memory, disk, or I/O impact is negligible.
427
+
428
+
###### Can enabling / using this feature result in resource exhaustion of some node resources (PIDs, sockets, inodes, etc.)?
418
429
419
-
DDOSing the API can lead to resource exhaustion.
430
+
No, because the endpoint queries existing data structures inside the kubelet.
420
431
421
432
### Troubleshooting
422
433
423
434
###### How does this feature react if the API server and/or etcd is unavailable?
424
435
425
-
N/A.
436
+
No impact, the feature is node-local.
426
437
427
438
###### What are other known failure modes?
428
439
429
-
The API will always return a well-known error. In normal operation, the API is expected to never return an error and always return a valid response, because it utilizes internal kubelet data which is always available. Bugs may cause the API to return unexpected errors, or to return inconsistent data. Consumers of the API should treat unexpected errors as bugs of this API.
440
+
feature gate disabled: The API will always return a well-known error. In normal operation, the API is expected to never return an error and always return a valid response, because it utilizes internal kubelet data which is always available.
441
+
Bugs may cause the API to return unexpected errors, or to return inconsistent data.
442
+
Consumers of the API should treat unexpected errors as bugs of this API.
430
443
431
444
###### What steps should be taken if SLOs are not being met to determine the problem?
432
445
433
-
N/A.
446
+
Check the error code to learn if the consumer of the API is being throttle by rate limiting introduced in the parent KEP 606.
447
+
Check the kubelet logs to learn about resource allocation errors.
434
448
435
449
## Implementation History
436
450
437
451
- 2023-01-12: KEP created
438
452
439
453
- 2024-09-10: KEP Updated to reflect the current state of the implementation.
0 commit comments