Automatic merge from submit-queue (batch tested with PRs 68119, 68191). If you want to cherry-pick this change to another branch, please follow the instructions here: https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md.
fix token controller keyFunc bug
Currently, token manager use keyFunc like: `fmt.Sprintf("%q/%q/%#v", name, namespace, tr.Spec)`.
Since tr.Spec contains point fields, new token request would not reuse the cache at all.
This patch fix this, also adds unit test.
```release-note
NONE
```
Automatic merge from submit-queue (batch tested with PRs 68087, 68256, 64621, 68299, 68296). If you want to cherry-pick this change to another branch, please follow the instructions here: https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md.
Change plugin watcher registration mechanism
**Which issue(s) this PR fixes**: #64637
**Notes For Reviewers**:
The current API the plugin watcher exposes to kubelet is the following:
```golang
type RegisterCallbackFn func(pluginName string, endpoint string,
versions []string, socketPath string) (error, chan bool)
```
The callback channel is here to signal the plugin watcher consumer when the plugin watcher API has notified the plugin of it's successful registration.
In other words the current lifecycle of a plugin is the following:
```
(pluginwatcher) GetInfo -> (pluginwatcher) NotifyRegistrationStatus -> (deviceplugin) ListWatch
```
Rather than
```
(pluginwatcher) GetInfo (race) -> (pluginwatcher) NotifyRegistrationStatus
(race) -> (deviceplugin) ListWatch
```
This PR changes the callback/channel mechanism to a more explicit, interfaced based contract (and more maintainable than a function to which we add more channels for more lifecycle events).
This PR also introduces three new states: {Init, Register, DeRegister}
```golang
// PluginHandler is an interface a client of the pluginwatcher API needs to implement in
// order to consume plugins
// The PluginHandler follows the simple following state machine:
//
// +--------------------------------------+
// | ReRegistration |
// | Socket created with same plugin name |
// | |
// | |
// Socket Created v + Socket Deleted
// +------------------> Validate +----------> Init +---------> Register +------------------> DeRegister
// + + +
// | | |
// | Error | Error |
// | | |
// v v v
// Out Out Out
//
// The pluginwatcher module follows strictly and sequentially this state machine for each *plugin name*.
// e.g: If you are Registering a plugin foo, you cannot get a DeRegister call for plugin foo
// until the Register("foo") call returns. Nor will you get a Validate("foo", "Different endpoint", ...)
// call until the Register("foo") call returns.
//
// ReRegistration: Socket created with same plugin name, usually for a plugin update
// e.g: plugin with name foo registers at foo.com/foo-1.9.7 later a plugin with name foo
// registers at foo.com/foo-1.9.9
//
// DeRegistration: When ReRegistration happens only the deletion of the new socket will trigger a DeRegister call
type PluginHandler interface {
// Validate returns an error if the information provided by
// the potential plugin is erroneous (unsupported version, ...)
ValidatePlugin(pluginName string, endpoint string, versions []string) error
// Init starts the plugin (e.g: contact the gRPC client, gets plugin
// specific information, ...) but if another plugin with the same name
// exists does not switch to the newer one.
// Any error encountered here can still be Notified to the plugin.
InitPlugin(pluginName string, endpoint string) error
// Register is called once the pluginwatcher has notified the plugin
// of its successful registration.
// Errors at this point can no longer be bubbled up to the plugin
RegisterPlugin(pluginName, endpoint string)
// DeRegister is called once the pluginwatcher observes that the socket has
// been deleted.
DeRegisterPlugin(pluginName string)
}
```
```release-note
NONE
```
/sig node
/area hw-accelerators
/cc @jiayingz @vikaschoudhary16 @vishh @vladimirvivien @sbezverk @figo (ccing the main reviewers of the original PR, feel free to cc more people)
Automatic merge from submit-queue. If you want to cherry-pick this change to another branch, please follow the instructions here: https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md.
Let the service controller retry when presistUpdate returns a conflict error
**What this PR does / why we need it**:
If a load balancer is changed while provisioning, it will fall into an error state and will not self-recover.
This PR picks up the conflict error and let serviceController retry in order to get the load balancer out of error state.
**Special notes for your reviewer**:
/assign @MrHohn @rramkumar1
**Release note**:
```release-note
Let service controller retry creating load balancer when persistUpdate failed due to conflict.
```
Since tr.Spec contains point fields, new token request would not reuse
the cache at all. This patch fix this, also adds unit test.
Signed-off-by: Mike Danese <mikedanese@google.com>
Automatic merge from submit-queue. If you want to cherry-pick this change to another branch, please follow the instructions here: https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md.
Resolves#59015, Scheduler: Add support for EBS types t3, r5, & z1d
Fixes#59015
The new t3, r5, r5d and z1 need matched as well according to this:
From current AWS documentation:
https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/volume_limits.html
T3, C5, C5d, M5, M5d, R5, R5d, and z1d instances support a maximum of
28 attachments, and every instance has at least one network interface
attachment. If you have no additional network interface attachments on
these instances, you could attach 27 EBS volumes.
**Release note**:
```NONE
```
Automatic merge from submit-queue. If you want to cherry-pick this change to another branch, please follow the instructions here: https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md.
Kubelet: only sync iptables on linux
**What this PR does / why we need it**:
Iptables is only supported on Linux, kubelet should only sync NAT rules on Linux.
Without this PR, Kubelet on Windows would logs following errors on each `syncNetworkUtil()`:
```
kubelet.err.log:4692:E0711 22:03:42.103939 2872 kubelet_network.go:102] Failed to ensure that nat chain KUBE-MARK-DROP exists: error creating chain "KUBE-MARK-DROP": executable file
```
**Which issue(s) this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close the issue(s) when PR gets merged)*:
Fixes#65713
**Special notes for your reviewer**:
**Release note**:
```release-note
Kubelet now only sync iptables on Linux.
```
Automatic merge from submit-queue. If you want to cherry-pick this change to another branch, please follow the instructions here: https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md.
Update bazel deps
**What this PR does / why we need it**: updates relevant bazel dependencies to pull in minor feature enhancements and bug fixes:
* rules_go 0.15.3: some speed improvements and better compatibility on Windows and Mac, among other bugfixes / enhancements
* repo-infra: fixes needed to support rules_go 0.15.1, and also a fix for https://github.com/kubernetes/kubernetes/pull/65501#issuecomment-400761696
* rules_docker v0.5.1: various enhancements/fixes, but we need this to support new manifest lists from upstream
* busybox latest: we last updated this a while ago. since Docker hub now uses manifest lists for core images like this, we needed the rules_docker update
* gazelle 0.14.0: among other enhancements, adds a feature to remove the need to build protoc for every build, which we aren't even using.
**Special notes for your reviewer**:
Depends on https://github.com/kubernetes/repo-infra/pull/81. I believe we also need bazel 0.16+.
**Release note**:
```release-note
NONE
```
Automatic merge from submit-queue (batch tested with PRs 68171, 67945, 68233). If you want to cherry-pick this change to another branch, please follow the instructions here: https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md.
Move the CloudControllerManagerConfiguration to an API group in `cmd/`
**What this PR does / why we need it**:
This PR is the last piece of https://github.com/kubernetes/kubernetes/issues/67233.
It moves the `CloudControllerManagerConfiguration` to its own `cloudcontrollermanager.config.k8s.io` config API group, but unlike the other components this API group is "private" (only available in `k8s.io/kubernetes`, which limits consumer base), as it's located entirely in `cmd/` vs a staging repo.
This decision was made for now as we're not sure what the story for the ccm loading ComponentConfig files is, and probably a "real" file-loading ccm will never exist in core, only helper libraries. Eventually the ccm will only be a library in any case, and implementors will/can use the base types the ccm library API group provides. It's probably good to note that there is no practical implication of this change as the ccm **cannot** read ComponentConfig files. Hencec the code move isn't user-facing.
With this change, we're able to remove `pkg/apis/componentconfig`, as this was the last consumer. That is hence done in this PR as well (so the move is easily visible in git, vs first one "big add" then a "big remove"). The only piece of code that was used was the flag helper structs, so I moved them to `pkg/util/flag` that I think makes sense for now.
**Which issue(s) this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close the issue(s) when PR gets merged)*:
ref: kubernetes/community#2354
**Special notes for your reviewer**:
This PR builds on top of (first two commits, marked as `Co-authored by: @stewart-yu`) https://github.com/kubernetes/kubernetes/pull/67689
**Release note**:
```release-note
NONE
```
/assign @liggitt @sttts @thockin @stewart-yu
From current AWS documentation:
https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/volume_limits.html
T3, C5, C5d, M5, M5d, R5, R5d, and z1d instances support a maximum of
28 attachments, and every instance has at least one network interface
attachment. If you have no additional network interface attachments on
these instances, you could attach 27 EBS volumes.
Automatic merge from submit-queue (batch tested with PRs 68161, 68023, 67909, 67955, 67731). If you want to cherry-pick this change to another branch, please follow the instructions here: https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md.
CSI: skip attach for non-attachable drivers
**What this PR does / why we need it**:
This is implementation of https://github.com/kubernetes/community/pull/2523. CSI volumes that don't need attach/detach now don't need external attacher running.
WIP:
* contains #67803 to get CSIDriver API. Ignore the first commit.
* ~~missing e2e test~~
/sig storage
cc: @saad-ali @vladimirvivien @verult @msau42 @gnufied @davidz627
**Release note**:
```release-note
CSI volume plugin does not need external attacher for non-attachable CSI volumes.
```
Automatic merge from submit-queue. If you want to cherry-pick this change to another branch, please follow the instructions here: https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md.
fix typo in ipvs RADME.md
**What this PR does / why we need it**:
fix typos in ipvs README.md
**Which issue(s) this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close the issue(s) when PR gets merged)*:
Fixes #
**Special notes for your reviewer**:
no
**Release note**:
```release-note
```
Automatic merge from submit-queue. If you want to cherry-pick this change to another branch, please follow the instructions here: https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md.
Implement semantic comparison of VolumeNodeAffinity for unit tests
**What this PR does / why we need it**:
Implements a semantic comparison function for VolumeNodeAffinity that is not sensitive to ordering of various members. Previous reflect.DeepEqual was sensitive to ordering causing it to be flaky.
**Which issue(s) this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close the issue(s) when PR gets merged)*:
Fixes # https://github.com/kubernetes/kubernetes/issues/67852
**Special notes for your reviewer**:
We want to able to successfully match `VolumeNodeAffinity{Required:&NodeSelector{NodeSelectorTerms:[{[{a In [1]} {b In [2 3]}] []}],},}` and `VolumeNodeAffinity{Required:&NodeSelector{NodeSelectorTerms:[{[{b In [3 2]} {a In [1]}] []}],},}` without being sensitive to the ordering of requirements with key `a` and `b` or the order of values with key `b`. This fix enables such semantic comparison of VolumeNodeAffinity
We can move `volumeNodeAffinitiesEqual` to volume/utils post code freeze
**Release note**:
```release-note
NONE
```
/sig storage
cc @msau42
Automatic merge from submit-queue. If you want to cherry-pick this change to another branch, please follow the instructions here: https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md.
Deprecation notice of the storage-versions flag
This PR deprecates the `--storage-versions` flag of kube-apiserver. The flag allows user to change the API version the content in etcd will be serialized to.
The flag provides unnecessary flexibility, with the side-effect of making kube-apiserver upgrades/downgrades hard to reason about. Specifically, Kubernetes follows the 4-step API version deprecation [policy](https://kubernetes.io/docs/reference/using-api/deprecation-policy/), which makes upgrades/downgrades across minor versions safe as long as all the data in etcd is encoded to the default storage versions. However, if users can specify their own storage versions, then the safety of each upgrade/downgrade needs to be analyzed case-by-case.
```release-note
Action required: The --storage-versions flag of kube-apiserver is deprecated. Please omit this flag to ensure the default storage versions are used. Otherwise the cluster is not safe to upgrade to a version newer than 1.12. This flag will be removed in 1.13.
```
/assign @deads2k @liggitt @lavalamp @enj
/sig api-machinery
Automatic merge from submit-queue (batch tested with PRs 67709, 67556). If you want to cherry-pick this change to another branch, please follow the instructions here: https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md.
Fix volume scheduling issue with pod affinity and anti-affinity
**What this PR does / why we need it**:
The previous design of the volume scheduler had volume assume + bind done before pod assume + bind. This causes issues when trying to evaluate future pods with pod affinity/anti-affinity because the pod has not been assumed while the volumes have been decided.
This PR changes the design so that volume and pod are assumed first, followed by volume and pod binding. Volume binding waits (asynchronously) for the operations to complete or error. This eliminates the subsequent passes through the scheduler to wait for volume binding to complete (although pod events or resyncs may still cause the pod to run through scheduling while binding is still in progress). This design also aligns better with the scheduler framework design, so will make it easier to migrate in the future.
Many changes had to be made in the volume scheduler to handle this new design, mostly around:
* How we cache pending binding operations. Now, any delayed binding PVC that is not fully bound must have a cached binding operation. This also means bind API updates may be repeated.
* Waiting for the bind operation to fully complete, and detecting failure conditions to abort the bind and retry scheduling.
**Which issue(s) this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close the issue(s) when PR gets merged)*:
Fixes#65131
**Special notes for your reviewer**:
**Release note**:
```release-note
Fixes issue where pod scheduling may fail when using local PVs and pod affinity and anti-affinity without the default StatefulSet OrderedReady pod management policy
```
Automatic merge from submit-queue. If you want to cherry-pick this change to another branch, please follow the instructions here: https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md.
Kubelet: only apply default hard evictions of nodefs.inodesFree on Linux
**What this PR does / why we need it**:
Kubelet sets default hard evictions of `nodefs.inodesFree ` for all platforms today. This will cause errors on Windows and a lot `no observation found for eviction signal nodefs.inodesFree` errors will be logs for kubelet.
```
kubelet.err.log:4961:W0711 22:21:12.378789 2872 helpers.go:808] eviction manager: no observation found for eviction signal nodefs.inodesFree
kubelet.err.log:4967:W0711 22:21:30.411371 2872 helpers.go:808] eviction manager: no observation found for eviction signal nodefs.inodesFree
kubelet.err.log:4974:W0711 22:21:48.446456 2872 helpers.go:808] eviction manager: no observation found for eviction signal nodefs.inodesFree
kubelet.err.log:4978:W0711 22:22:06.482441 2872 helpers.go:808] eviction manager: no observation found for eviction signal nodefs.inodesFree
```
This PR updates the default hard eviction value and only apply nodefs.inodesFree on Linux.
**Which issue(s) this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close the issue(s) when PR gets merged)*:
Fixes#66088
**Special notes for your reviewer**:
**Release note**:
```release-note
Kubelet only applies default hard evictions of nodefs.inodesFree on Linux
```
Automatic merge from submit-queue. If you want to cherry-pick this change to another branch, please follow the instructions here: https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md.
Add kubelet stats for windows system container "pods"
**What this PR does / why we need it**:
This PR adds kubelet stats for windows system container "pods". Without this, kubelet will always logs error:
```
kubelet.err.log:4832:E0711 22:12:49.241358 2872 helpers.go:735] eviction manager: failed to construct signal: "allocatableMemory.available" error: system container "pods" not found
```
**Which issue(s) this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close the issue(s) when PR gets merged)*:
Fixes#66087
**Special notes for your reviewer**:
/sig windows
/sig node
**Release note**:
```release-note
Add kubelet stats for windows system container "pods"
```
Automatic merge from submit-queue (batch tested with PRs 66840, 68159). If you want to cherry-pick this change to another branch, please follow the instructions here: https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md.
CSI Cluster Registry and Node Info CRDs Improvements
**What this PR does / why we need it**:
https://github.com/kubernetes/kubernetes/pull/67803 merged before I could address @lavalamp's feedback. This PR addresses his feedback
**Which issue(s) this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close the issue(s) when PR gets merged)*:
Follow up on PR https://github.com/kubernetes/kubernetes/pull/67803
**Special notes for your reviewer**:
**Release note**:
```release-note
```
/assign @lavalamp
/assign @thockin
CC @jsafrane @vladimirvivien @verult @gnufied @childsb
Automatic merge from submit-queue (batch tested with PRs 66840, 68159). If you want to cherry-pick this change to another branch, please follow the instructions here: https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md.
TTL for cleaning up Jobs after they finish
**What this PR does / why we need it**: https://github.com/kubernetes/features/issues/592
**Which issue(s) this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close the issue(s) when PR gets merged)*:
Fixes#64470
For https://github.com/kubernetes/features/issues/592
**Special notes for your reviewer**: @kubernetes/sig-apps-pr-reviews
**Release note**:
```release-note
Add a TTL machenism to clean up Jobs after they finish.
```
* FindPodVolumes
* Prebound PVCs are treated like unbound immediate PVCs and will error
* Always check for fully bound PVCs and cache bindings for not fully
bound PVCs
* BindPodVolumes
* Retry API updates for not fully bound PVCs even if the assume cache
already marked it
* Wait for PVCs to be fully bound after making the API updates
* Error when detecting binding/provisioning failure conditions
