Automatic merge from submit-queue
Retry scheduling pods after errors more consistently in scheduler
**What this PR does / why we need it**:
This fixes 2 places in the scheduler where pods can get stuck in Pending forever. In both these places, errors happen and `sched.config.Error` is not called afterwards. This is a problem because `sched.config.Error` is responsible for requeuing pods to retry scheduling when there are issues (see [here](2540b333b2/plugin/pkg/scheduler/factory/factory.go (L958))), so if we don't call `sched.config.Error` then the pod will never get scheduled (unless the scheduler is restarted).
One of these (where it returns when `ForgetPod` fails instead of continuing and reporting an error) is a regression from [this refactor](https://github.com/kubernetes/kubernetes/commit/ecb962e6585#diff-67f2b61521299ca8d8687b0933bbfb19L234), and with the [old behavior](80f26fa8a8/plugin/pkg/scheduler/scheduler.go (L233-L237)) the error was reported correctly. As far as I can tell changing the error handling in that refactor wasn't intentional.
When AssumePod fails there's never been an error reported but I think adding this will help the scheduler recover when something goes wrong instead of letting pods possibly never get scheduled.
This will help prevent issues like https://github.com/kubernetes/kubernetes/issues/49314 in the future.
**Release note**:
```release-note
Fix incorrect retry logic in scheduler
```
Automatic merge from submit-queue (batch tested with PRs 50119, 48366, 47181, 41611, 49547)
Task 0: Added node taints labels and feature flags
**What this PR does / why we need it**:
Added node taint const for node condition.
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: part of #42001
**Release note**:
```release-note
None
```
Automatic merge from submit-queue
Fix incorrect call to 'bind' in scheduler
I previously submitted https://github.com/kubernetes/kubernetes/pull/49661 -- I'm not sure if that PR is too big or what, but this is an attempt at a smaller PR that makes progress on the same issue and is easier to review.
**What this PR does / why we need it**:
In this refactor (https://github.com/kubernetes/kubernetes/commit/ecb962e6585#diff-67f2b61521299ca8d8687b0933bbfb19R223) the scheduler code was refactored into separate `bind` and `assume` functions. When that happened, `bind` was called with `pod` as an argument. The argument to `bind` should be the assumed pod, not the original pod. Evidence that `assumedPod` is the correct argument bind and not `pod`: 80f26fa8a8/plugin/pkg/scheduler/scheduler.go (L229-L234). (and it says `assumed` in the function signature for `bind`, even though it's not called with the assumed pod as an argument).
This is an issue (and causes #49314, where pods that fail to bind to a node get stuck indefinitely) in the following scenario:
1. The pod fails to bind to the node
2. `bind` calls `ForgetPod` with the `pod` argument
3. since `ForgetPod` is expecting the assumed pod as an argument (because that's what's in the scheduler cache), it fails with an error like `scheduler cache ForgetPod failed: pod test-677550-rc-edit-namespace/nginx-jvn09 state was assumed on a different node`
4. The pod gets lost forever because of some incomplete error handling (which I haven't addressed here in the interest of making a simpler PR)
In this PR I've fixed the call to `bind` and modified the tests to make sure that `ForgetPod` gets called with the correct argument (the assumed pod) when binding fails.
**Which issue this PR fixes**: fixes#49314
**Special notes for your reviewer**:
**Release note**:
```release-note
```
Automatic merge from submit-queue (batch tested with PRs 48976, 49474, 40050, 49426, 49430)
Remove duplicated import and wrong alias name of api package
**What this PR does / why we need it**:
**Which issue this PR fixes**: fixes#48975
**Special notes for your reviewer**:
/assign @caesarxuchao
**Release note**:
```release-note
NONE
```
Automatic merge from submit-queue (batch tested with PRs 48636, 49088, 49251, 49417, 49494)
Fix issues for local storage allocatable feature
This PR fixes the following issues:
1. Use ResourceStorageScratch instead of ResourceStorage API to represent
local storage capacity
2. In eviction manager, use container manager instead of node provider
(kubelet) to retrieve the node capacity and reserved resources. Node
provider (kubelet) has a feature gate so that storagescratch information
may not be exposed if feature gate is not set. On the other hand,
container manager has all the capacity and allocatable resource
information.
This PR fixes issue #47809
Automatic merge from submit-queue (batch tested with PRs 48043, 48200, 49139, 36238, 49130)
Implement equivalence cache by caching and re-using predicate result
The last part of #30844, I opened a new PR instead of overwrite the old one because we changed some basic assumption by allowing invalidating equivalence cache item by individual predicate.
The idea of this PR is based on discussion in https://github.com/kubernetes/kubernetes/issues/32024
- [x] Pods belong to same controllerRef considered to be equivalent
- [x] ` podFitsOnNode` will use cached predicate result if it's available
- [x] Equivalence cache will be updated when if a fresh new predicate is done
- [x] `factory.go` will invalid specific predicate cache(s) based on the object change
- [x] Since `schedule` and `bind` are async, we need to optimistically invalid affected cache(s) before `bind`
- [x] Fully unit test of affected files
- [x] e2e test to verify cache update/invalid workflow
- [x] performance test results
- [x] Some nits fixes related but expected to result in `needs-rebase` so they are split to: #36060#35968#37512
cc @wojtek-t @davidopp
Automatic merge from submit-queue (batch tested with PRs 49055, 49128, 49132, 49134, 49110)
Remove affinity annotations leftover
**What this PR does / why we need it**:
This is a further cleanup for affinity annotations, following #47869.
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: fixes #
ref: #47869
**Special notes for your reviewer**:
- I remove the commented test cases and just leave TODOs instead. I think converting these untestable test cases for now is not necessary. We can add new test cases in future.
- I remove the e2e test case `validates that embedding the JSON PodAffinity and PodAntiAffinity setting as a string in the annotation value work` because we have a test case `validates that InterPod Affinity and AntiAffinity is respected if matching` to test the same thing.
/cc @aveshagarwal @bsalamat @gyliu513 @k82cn @timothysc
**Release note**:
```release-note
NONE
```
Automatic merge from submit-queue (batch tested with PRs 46094, 48544, 48807, 49102, 44174)
Static deepcopy – phase 1
This PR is the follow-up of https://github.com/kubernetes/kubernetes/pull/36412, replacing the
dynamic reflection based deepcopy with static DeepCopy+DeepCopyInto methods on API types.
This PR **does not yet** include the code dropping the cloner from the scheme and all the
porting of the calls to scheme.Copy. This will be part of a follow-up "Phase 2" PR.
A couple of the commits will go in first:
- [x] audit: fix deepcopy registration https://github.com/kubernetes/kubernetes/pull/48599
- [x] apimachinery+apiserver: separate test types in their own packages #48601
- [x] client-go: remove TPR example #48604
- [x] apimachinery: remove unneeded GetObjectKind() impls #48608
- [x] sanity check against origin, that OpenShift's types are fine for static deepcopy https://github.com/deads2k/origin/pull/34
TODO **after** review here:
- [x] merge https://github.com/kubernetes/gengo/pull/32 and update vendoring commit
Automatic merge from submit-queue (batch tested with PRs 48333, 48806, 49046)
use v1.ResourcePods instead of hard coding "pods"
Signed-off-by: sakeven <jc5930@sina.cn>
**What this PR does / why we need it**:
use v1.ResourcePods instead of hard coding 'pods'
**Special notes for your reviewer**:
**Release note**:
```
NONE
```
Automatic merge from submit-queue
[Scheduler] Remove error since err is always nil
Signed-off-by: sakeven <jc5930@sina.cn>
**What this PR does / why we need it**:
No need to log error since err is always nil.
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: fixes #
**Special notes for your reviewer**:
**Release note**:
```
NONE
```
Automatic merge from submit-queue (batch tested with PRs 48262, 48805)
[Scheduler] Use const value maxPriority instead of immediate value 10
Signed-off-by: sakeven <jc5930@sina.cn>
**What this PR does / why we need it**:
Use const value maxPriority instead of immediate value 10.
**Special notes for your reviewer**:
**Release note**:
```
NONE
```
Automatic merge from submit-queue
forget pod first after binding failed
Signed-off-by: sakeven <jc5930@sina.cn>
**What this PR does / why we need it**:
In the implementation of scheduler cache, `FinishBinding` marks Pod expired, and then pod would be cleaned in ttl seconds. While `ForgetPod` checks Pod whether assumed, if not, it reports an error.
So if binding failed and ttl(now 30s) is too short, the error will occur when `ForgetPod`, thus we won't record `BindingRejected` event.
Although it's rare, we shouldn't depend on the value of ttl.
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: fixes #
**Special notes for your reviewer**:
**Release note**:
```
NONE
```
Automatic merge from submit-queue (batch tested with PRs 47417, 47638, 46930)
Added scheduler integration test owners.
**What this PR does / why we need it**:
Add OWNER file into scheduler integration test.
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: fixes # N/A
**Release note**:
```release-note-none
```
Automatic merge from submit-queue
Improved code coverage for equivalence cache.
**What this PR does / why we need it**:
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: fixes #
**Special notes for your reviewer**:
**Release note**:
```release-note
none
```
This PR fixes the following issues:
1. Use ResourceStorageScratch instead of ResourceStorage API to represent
local storage capacity
2. In eviction manager, use container manager instead of node provider
(kubelet) to retrieve the node capacity and reserved resources. Node
provider (kubelet) has a feature gate so that storagescratch information
may not be exposed if feature gate is not set. On the other hand,
container manager has all the capacity and allocatable resource
information.
Automatic merge from submit-queue (batch tested with PRs 48405, 48742, 48748, 48571, 48482)
Removed scheduler dependencies to testapi.
**What this PR does / why we need it**:
When refactor scheduler to use client-go, k8s.io/api, it's also need to remove the dependeny to testapi.
prefer to only include import/BUILD changes for #44188, so created separated PR for other enhancement removal.
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: partially fixes#44188
**Release note**:
```release-note-none
```
Automatic merge from submit-queue (batch tested with PRs 46865, 48661, 48598, 48658, 48614)
Fix function names in the comments
This patch fixes function and type names in the comments
in predicates.go.
**What this PR does / why we need it**:
It fixes function and type names in the comments in predicates.go.
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: fixes #
This does not have an issue # because it is a trivial fix.
**Special notes for your reviewer**:
**Release note**:
```release-note
```
Automatic merge from submit-queue (batch tested with PRs 48402, 47203, 47460, 48335, 48322)
HTTPExtender: should close resp.Body even when StatusCode not ok
Signed-off-by: sakeven <jc5930@sina.cn>
**What this PR does / why we need it**:
close resp.Body even when StatusCode isn't ok
**Special notes for your reviewer**:
**Release note**:
```
NONE
```
Automatic merge from submit-queue
scheduler: fix validation test
Signed-off-by: sakeven <jc5930@sina.cn>
**What this PR does / why we need it**:
Without setting `Weight`, `ValidatePolicy` will report
```
Priority for extender http://127.0.0.1:8081/extender should have a positive weight applied to it
```
**Besides**, it seems it's not a good way to test ValidatePolicy by```if ValidatePolicy(extenderPolicy) == nil```, because we can't determine specific reason which causes error.
**Special notes for your reviewer**:
**Release note**:
```
NONE
```
Automatic merge from submit-queue (batch tested with PRs 48399, 48450, 48144)
Group and order imported packages.
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: fixes #N/A
**Release note**:
```release-note-none
```
Automatic merge from submit-queue (batch tested with PRs 47043, 48448, 47515, 48446)
Refactor slice intersection
**What this PR does / why we need it**:
In worst case, the original method is O(N^2), while current method is 3 * O(N).
I think it is better.
**Release note**:
```release-note
NONE
```
Automatic merge from submit-queue (batch tested with PRs 46926, 48468)
Added helper funcs to schedulercache.Resource.
**What this PR does / why we need it**:
Avoid duplicated code slice by helper funcs.
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: fixes#46924
**Release note**:
```release-note-none
```
Automatic merge from submit-queue
Cleanup predicates.go
**What this PR does / why we need it**:
cleanup some comments and errors.New().
**Special notes for your reviewer**:
/cc @jayunit100
**Release note**:
```release-note
NONE
```
Automatic merge from submit-queue
Removes alpha feature gate for affinity annotations.
**What this PR does / why we need it**:
In 1.5 we added a backstop to support alpha affinity annotations. This PR removes that support in favor of the Beta fields per discussions.
It also serves as a precursor to some of the component config work that @ncdc has done around @mikedanese design proposal.
xref: https://github.com/kubernetes/kubernetes/pull/41617
**Special notes for your reviewer**:
**Release note**:
```
Removes alpha feature gate for pod affinity annotations.
```
/cc @kubernetes/sig-scheduling-pr-reviews @kubernetes/sig-cluster-lifecycle-misc
Automatic merge from submit-queue (batch tested with PRs 47883, 47179, 46966, 47982, 47945)
Fix local isolation for pod requesting only overlay or scratch
**What this PR does / why we need it**:
Fix overlay resource predicates for pod with only overlay or scratch storage request.
E.g. the following pod can pass predicate even if overlay is only 512Gi.
```yaml
apiVersion: v1
kind: Pod
metadata:
name: pod
spec:
containers:
- name: nginx
image: nginx
resources:
requests:
storage.kubernetes.io/overlay: 1024Gi
```
similarly, following pod will also pass predicate
```yaml
apiVersion: v1
kind: Pod
metadata:
name: pod
spec:
containers:
- name: nginx
image: nginx
volumeMounts:
- name: data
mountPath: /data
volumes:
- name: data
emptyDir:
sizeLimit: 1024Gi
```
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: fixes https://github.com/kubernetes/kubernetes/issues/47798
**Special notes for your reviewer**:
**Release note**:
```release-note
```
@jingxu97 @vishh @dashpole
Automatic merge from submit-queue (batch tested with PRs 45877, 46846, 46630, 46087, 47003)
Remove duplicate errors from an aggregate error input.
This PR, in general, removes duplicate errors from an aggregate error input, and returns unique errors with their occurrence count. Specifically, this PR helps with some scheduler errors that fill the log enormously. For example, see the following `truncated` output from a 300-plus nodes cluster, as there was a same error from almost all nodes.
[SchedulerPredicates failed due to persistentvolumeclaims "mongodb" not found, which is unexpected., SchedulerPredicates failed due to persistentvolumeclaims "mongodb" not found, which is unexpected., SchedulerPredicates failed due to persistentvolumeclaims "mongodb" not found, which is unexpected., SchedulerPredicates failed due to persistentvolumeclaims "mongodb" not found, which is unexpected., SchedulerPredicates failed due to persistentvolumeclaims "mongodb" not found, which is unexpected., SchedulerPredicates failed due to persistentvolumeclaims "mongodb" not found, which is unexpected., SchedulerPredicates failed due to persistentvolumeclaims "mongodb" not found, which is unexpected., SchedulerPredicates failed due to persistentvolumeclaims "mongodb" not found, which is unexpected., SchedulerPredicates failed due to persistentvolumeclaims "mongodb" not found, which is unexpected., SchedulerPredicates failed due to persistentvolumeclaims "mongodb" not found, which is unexpected., SchedulerPredicates failed due to persistentvolumeclaims "mongodb" not found, which is unexpected., SchedulerPredicates failed due to persistentvolumeclaims "mongodb" not found, which is unexpected., SchedulerPredicates failed due to persistentvolumeclaims "mongodb" not found, which is unexpected., SchedulerPredicates failed due to persistentvolumeclaims "mongodb" not found, which is unexpected., SchedulerPredicates failed due to persistentvolumeclaims "mongodb" not found, which is unexpected., SchedulerPredicates failed due to persistentvolumeclaims "mongodb" not found, which is unexpected., SchedulerPredicates failed due to persistentvolumeclaims "mongodb" not found, which is unexpected., SchedulerPredicates failed due to persistentvolumeclaims "mongodb" not found.........
After this PR, the output looks like (on a 2-node cluster):
SchedulerPredicates failed due to persistentvolumeclaims "mongodb" not found, which is unexpected.(Count=2)
@derekwaynecarr @smarterclayton @kubernetes/sig-scheduling-pr-reviews
Fixes https://github.com/kubernetes/kubernetes/issues/47145
Automatic merge from submit-queue (batch tested with PRs 47083, 44115, 46881, 47082, 46577)
Scheduler should not log an error when there is no fit
**What this PR does / why we need it**:
The scheduler should not log an error when it is unable to find a fit for a pod as it's an expected situation when resources are unavailable on the cluster that satisfy the pods requirements.
Automatic merge from submit-queue (batch tested with PRs 46787, 46876, 46621, 46907, 46819)
Highlight nodeSelector when checking nodeSelector for Pod.
**What this PR does / why we need it**:
Currently, we are using function name as `PodSelectorMatches` to check if `nodeSelector` matches for a Pod, it is better update the function name a bit to reflect it is checking `nodeSelector` for a Pod.
The proposal is rename `PodSelectorMatches` as `PodMatchNodeSelector`.
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: fixes #
**Special notes for your reviewer**:
**Release note**:
```release-note
NONE
```
Automatic merge from submit-queue
Add local storage (scratch space) allocatable support
This PR adds the support for allocatable local storage (scratch space).
This feature is only for root file system which is shared by kubernetes
componenets, users' containers and/or images. User could use
--kube-reserved flag to reserve the storage for kube system components.
If the allocatable storage for user's pods is used up, some pods will be
evicted to free the storage resource.
This feature is part of local storage capacity isolation and described in the proposal https://github.com/kubernetes/community/pull/306
**Release note**:
```release-note
This feature exposes local storage capacity for the primary partitions, and supports & enforces storage reservation in Node Allocatable
```
Automatic merge from submit-queue (batch tested with PRs 46239, 46627, 46346, 46388, 46524)
move labels to components which own the APIs
During the apimachinery split in 1.6, we accidentally moved several label APIs into apimachinery. They don't belong there, since the individual APIs are not general machinery concerns, but instead are the concern of particular components: most commonly the kubelet. This pull moves the labels into their owning components and out of API machinery.
@kubernetes/sig-api-machinery-misc @kubernetes/api-reviewers @kubernetes/api-approvers
@derekwaynecarr since most of these are related to the kubelet
This PR adds the check for local storage request when admitting pods. If
the local storage request exceeds the available resource, pod will be
rejected.
This PR adds the support for allocatable local storage (scratch space).
This feature is only for root file system which is shared by kubernetes
componenets, users' containers and/or images. User could use
--kube-reserved flag to reserve the storage for kube system components.
If the allocatable storage for user's pods is used up, some pods will be
evicted to free the storage resource.
Automatic merge from submit-queue (batch tested with PRs 46076, 43879, 44897, 46556, 46654)
Local storage plugin
**What this PR does / why we need it**:
Volume plugin implementation for local persistent volumes. Scheduler predicate will direct already-bound PVCs to the node that the local PV is at. PVC binding still happens independently.
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*:
Part of #43640
**Release note**:
```
Alpha feature: Local volume plugin allows local directories to be created and consumed as a Persistent Volume. These volumes have node affinity and pods will only be scheduled to the node that the volume is at.
```
Automatic merge from submit-queue
Move hardPodAffinitySymmetricWeight to scheduler policy config
**What this PR does / why we need it**:
Move hardPodAffinitySymmetricWeight to scheduler policy config
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: fixes#43845
**Special notes for your reviewer**:
If you like this, will add test later
**Release note**:
```
Move hardPodAffinitySymmetricWeight from KubeSchedulerConfiguration to scheduler Policy config
```
Automatic merge from submit-queue
removing generic_scheduler todo after discussion (#46027)
**What this PR does / why we need it**:
**Which issue this PR fixes** #46027
**Special notes for your reviewer**: just a quick clean cc @wojtek-t
**Release note**:
```release-note
```
Automatic merge from submit-queue (batch tested with PRs 45766, 46223)
Scheduler should use a shared informer, and fix broken watch behavior for cached watches
Can be used either from a true shared informer or a local shared
informer created just for the scheduler.
Fixes a bug in the cache watcher where we were returning the "current" object from a watch event, not the historic event. This means that we broke behavior when introducing the watch cache. This may have API implications for filtering watch consumers - but on the other hand, it prevents clients filtering from seeing objects outside of their watch correctly, which can lead to other subtle bugs.
```release-note
The behavior of some watch calls to the server when filtering on fields was incorrect. If watching objects with a filter, when an update was made that no longer matched the filter a DELETE event was correctly sent. However, the object that was returned by that delete was not the (correct) version before the update, but instead, the newer version. That meant the new object was not matched by the filter. This was a regression from behavior between cached watches on the server side and uncached watches, and thus broke downstream API clients.
```
Automatic merge from submit-queue
Moved qos to api.helpers.
**What this PR does / why we need it**:
The `GetPodQoS` is also used by other components, e.g. kube-scheduler and it's not bound to kubelet; moved it to api helpers so client-go.
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: fixes #N/A
**Release note**:
```release-note-none
```
Automatic merge from submit-queue
Disambiguate schedule, assume, and bind in functions as well as in
Addresses my comments in #45972 about how these metrics need to be disambiguated.
- separates schedule, assume, and bind.
- renames variables like `dest` to be explicit.
- removes the logging statement to occur outside of the timed portion of the metric measurement.
Generally makes `sheduleOne` a happy function to read :)
