Automatic merge from submit-queue. If you want to cherry-pick this change to another branch, please follow the instructions <a href="https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md">here</a>.
Make AllocateResponse artifacts global across all devices per container in device plugin API
The current version of Device plugins returns artifacts (env vars, mounts, devices) per device, per container. This is not necessary and results in complex merging issues on kubelet side
There can still be a conflict if the artifacts returned by device plugin conflicts with the pod spec. In that case, I'd recommend failing pods in kubelet. This is yet to be addressed.
The go package name for device plugin apis is updated from `pkg/kubelet/apis/deviceplugin/v1alpha1` to `pkg/kubelet/apis/deviceplugin/v1alpha` (sub-version dropped) because we expect the alpha version to change until it graduates to beta and changing the go package everytime the actual alpha version changes is too tedious.
```release-note
Device plugin Alpha API no longer supports returning artifacts per device as part of AllocateResponse.
```
TODO:
- [x] Bump kubelet side API version
- [X] Post an updated device plugin image that works with the new API version
- [ ] Stabilize e2es (This PR needs to be merged since there is a dependency on the plugin side for vendoring)
There is no use case known for passing artifacts per device as it currently exists. The current API is also
complex to use for simple clients. Hence this PR creates a flat namespace where artifacts like environment variables
and mount points apply globally to all devices returned as part of AllocateResponse proto.
Signed-off-by: Vishnu kannan <vishnuk@google.com>
Revert "Merge pull request #51857 from kubernetes/revert-51307-kc-type-refactor"
This reverts commit 9d27d92420, reversing
changes made to 2e69d4e625.
See original: #51307
We punted this from 1.8 so it could go through an API review. The point
of this PR is that we are trying to stabilize the kubeletconfig API so
that we can move it out of alpha, and unblock features like Dynamic
Kubelet Config, Kubelet loading its initial config from a file instead
of flags, kubeadm and other install tools having a versioned API to rely
on, etc.
We shouldn't rev the version without both removing all the deprecated
junk from the KubeletConfiguration struct, and without (at least
temporarily) removing all of the fields that have "Experimental" in
their names. It wouldn't make sense to lock in to deprecated fields.
"Experimental" fields can be audited on a 1-by-1 basis after this PR,
and if found to be stable (or sufficiently alpha-gated), can be restored
to the KubeletConfiguration without the "Experimental" prefix.
Command line flag API remains the same. This allows ComponentConfig
structures (e.g. KubeletConfiguration) to express the map structure
behind feature gates in a natural way when written as JSON or YAML.
For example:
KubeletConfiguration Before:
```
apiVersion: kubeletconfig/v1alpha1
kind: KubeletConfiguration
featureGates: "DynamicKubeletConfig=true,Accelerators=true"
```
KubeletConfiguration After:
```
apiVersion: kubeletconfig/v1alpha1
kind: KubeletConfiguration
featureGates:
DynamicKubeletConfig: true
Accelerators: true
```
Automatic merge from submit-queue. If you want to cherry-pick this change to another branch, please follow the instructions <a href="https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md">here</a>.
enable to specific unconfined AppArmor profile
**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#52370
**Special notes for your reviewer**:
/assign @tallclair @liggitt
**Release note**:
```release-note
enable to specific unconfined AppArmor profile
```
Automatic merge from submit-queue (batch tested with PRs 44596, 52708, 53163, 53167, 52692). If you want to cherry-pick this change to another branch, please follow the instructions <a href="https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md">here</a>.
Fix --kube-reserved storage key name and add UTs for node allocatable reservation
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: part of #52463
**Special notes for your reviewer**:
**Release note**:
```release-note
NONE
```
/assign @jingxu97
Automatic merge from submit-queue (batch tested with PRs 52990, 53064, 52686, 52221, 53069). If you want to cherry-pick this change to another branch, please follow the instructions <a href="https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md">here</a>.
Align imagefs eviction defaults with image gc defaults
**What this PR does / why we need it**:
If a node is configured to use an imagefs for container storage, we should align the default imagefs eviction threshold with the default image-gc threshold. This PR updates the default imagesfs.available threshold to trigger when below 15% available space, which is same as default image-gc high threshold for 85%.
Fixes https://github.com/kubernetes/kubernetes/issues/53074
**Special notes for your reviewer**:
none, this only impacts nodes running an imagefs otherwise its ignored.
**Release note**:
```release-note
NONE
```
Automatic merge from submit-queue (batch tested with PRs 52990, 53064, 52686, 52221, 53069). If you want to cherry-pick this change to another branch, please follow the instructions <a href="https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md">here</a>.
CRI: Allow configuring stdout/stderr streams for Exec/Attach requests
Add stdout/stderr to exec and attach requests. Also check the request to
ensure it meets the requirements.
**Which issue this PR fixes**: fixes#44448
```release-note
CRI: Add stdout/stderr fields to Exec and Attach requests.
```
Automatic merge from submit-queue. If you want to cherry-pick this change to another branch, please follow the instructions <a href="https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md">here</a>.
Add an OWNERS file for deviceplugin package. Update OWNERS file for gpu package.
**Release note**:
```release-note
NONE
```
Modifies the VolumeZonePredicate to handle a PV that belongs to more
then one zone or region. This is indicated by the zone or region label
value containing a comma separated list.
Automatic merge from submit-queue (batch tested with PRs 49971, 51357, 51616, 51649, 51372)
CPU manager wiring and `none` policy
Blocker for CPU manager #49186 (4 of 6)
* Previous PR in this series: #51140
* Next PR in this series: #51180
cc @balajismaniam @derekwaynecarr @sjenning
**Release note**:
```release-note
NONE
```
TODO:
- [X] In-memory CPU manager state
- [x] Kubelet config value
- [x] Feature gate
- [X] None policy
- [X] Unit tests
- [X] CPU manager instantiation
- [x] Calls into CPU manager from Kubelet container runtime
Automatic merge from submit-queue (batch tested with PRs 51513, 51515, 50570, 51482, 51448)
Add PVCRef to VolumeStats
**What this PR does / why we need it**:
For pod volumes that reference a PVC, add a PVCRef to the corresponding
volume stat. This allows metrics to be indexed/queried by PVC name
which is more user-friendly than Pod reference
**Which issue this PR fixes** : [#363](https://github.com/kubernetes/features/issues/363)
**Special notes for your reviewer**:
**Release note**:
```
`VolumeStats` reported by the kubelet stats summary API
(http://<node>:10255/stats/summary) now include a PVCRef
field describing the PVC referenced by the volume (if any).
```
Automatic merge from submit-queue (batch tested with PRs 50719, 51216, 50212, 51408, 51381)
Use constants instead of magic string for runtime names
**What this PR does / why we need it**:
Use constants instead of magic string for runtime names.
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: fixes#51678
**Special notes for your reviewer**:
**Release note**:
```release-note
NONE
```
For pod volumes that reference a PVC, add a PVCRef to the corresponding
volume stat. This allows metrics to be indexed/queried by PVC name
which is more user-friendly than Pod reference
Automatic merge from submit-queue (batch tested with PRs 49961, 50005, 50738, 51045, 49927)
adding validations on kubelet starting configurations
**What this PR does / why we need it**:
I found some validations of kubelet starting options were missing when I was creating a custom cluster from scratch. The kubelet does not check invalid configurations on `--cadvisor-port`, `--event-burst`, `--image-gc-high-threshold`, etc. I have added some validations in kubelet like validations in `cmd/kube-apiserver/app/options/validation.go`.
**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
Adds additional validation for kubelet in `pkg/kubelet/apis/kubeletconfig/validation`.
```
Automatic merge from submit-queue (batch tested with PRs 50932, 49610, 51312, 51415, 50705)
Deprecation warnings for auto detecting cloud providers
**What this PR does / why we need it**:
Adds deprecation warnings for auto detecting cloud providers. As part of the initiative for out-of-tree cloud providers, this feature is conflicting since we're shifting the dependency of kubernetes core into cAdvisor. In the future kubelets should be using `--cloud-provider=external` or no cloud provider at all.
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: fixes#50986
**Special notes for your reviewer**:
NOTE: I still have to coordinate with sig-node and kubernetes-dev to get approval for this deprecation, I'm only opening this PR since we're close to code freeze and it's something presentable.
**Release note**:
```release-note
Deprecate auto detecting cloud providers in kubelet. Auto detecting cloud providers go against the initiative for out-of-tree cloud providers as we'll now depend on cAdvisor integrations with cloud providers instead of the core repo. In the near future, `--cloud-provider` for kubelet will either be an empty string or `external`.
```
Automatic merge from submit-queue
Remove null -> [] slice hack
Closes#44593
When 1.6 added protobuf storage, the storage layer lost the ability to persist slice fields with empty but non-null values.
As a workaround, we tried to convert empty slice fields to `[]`, rather than `null`. Compressing `null` -> `[]` was just as much of an API breakage as `[]` -> `null`, but was hoped to cause fewer problems in clients that don't do null checks.
Because of conversion optimizations around converting lists of objects, the `null` -> `[]` hack was discovered to only apply to individual get requests, not to a list of objects. 1.6 and 1.7 was released with this behavior, and the world didn't explode. 1.7 documented the breaking API change that `null` and `[]` should be considered equivalent, unless otherwise noted on a particular field.
This PR:
* Reverts the earlier attempt (https://github.com/kubernetes/kubernetes/pull/43422) at ensuring non-null json slice output in conversion
* Makes results of `get` consistent with the results of `list` (which helps naive clients that do deepequal comparisons of objects obtained via list/watch and get), and allows empty slice fields to be returned as `null`
```release-note
Protobuf serialization does not distinguish between `[]` and `null`.
API fields previously capable of storing and returning either `[]` and `null` via JSON API requests (for example, the Endpoints `subsets` field) can now store only `null` when created using the protobuf content-type or stored in etcd using protobuf serialization (the default in 1.6+). JSON API clients should tolerate `null` values for such fields, and treat `null` and `[]` as equivalent in meaning unless specifically documented otherwise for a particular field.
```
The ReadOnlyPort defaulting prevented passing 0 to diable via
the KubeletConfiguraiton struct.
The HealthzPort defaulting prevented passing 0 to disable via the
KubeletConfiguration struct. The documentation also failed to mention
this, but the check is performed in code.
The CAdvisorPort documentation failed to mention that you can pass 0 to
disable.
Automatic merge from submit-queue
Remove crash loop "detection" from the dynamic kubelet config feature
**What this PR does / why we need it**:
The subfeature was a cool idea, but in the end it is very complex to
separate Kubelet restarts into crash-loops caused by config vs.
crash-loops caused by other phenomena, like admin-triggered node restarts,
kernel panics, and and process babysitter behavior. Dynamic kubelet config
will be better off without the potential for false positives here.
Removing this subfeature also simplifies dynamic configuration by
reducing persistent state:
- we no longer need to track bad config in a file
- we no longer need to track kubelet startups in a file
**Which issue this PR fixes**: fixes#50216
**Release note**:
```release-note
NONE
```
Automatic merge from submit-queue (batch tested with PRs 50489, 51070, 51011, 51022, 51141)
Fixed code comments that were not updated
**What this PR does / why we need it**:
The comment of the args ‘KubeReserved’ is out of date and there is no consistent with command line messages
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: fixes #
**Release note**:
```
NONE
```
The subfeature was a cool idea, but in the end it is very complex to
separate Kubelet restarts into crash-loops caused by config vs.
crash-loops caused by other phenomena, like admin-triggered node restarts,
kernel panics, and and process babysitter behavior. Dynamic kubelet config
will be better off without the potential for false positives here.
Removing this subfeature also simplifies dynamic configuration by
reducing persistent state:
- we no longer need to track bad config in a file
- we no longer need to track kubelet startups in a file
This reverts commit f4afdecef8, reversing
changes made to e633a1604f.
This also fixes a bug where Kubemark was still using the core api scheme
to manipulate the Kubelet's types, which was the cause of the initial
revert.
Automatic merge from submit-queue (batch tested with PRs 49342, 50581, 50777)
Device Plugin Protobuf API
**What this PR does / why we need it:**
This implements the Device Plugin API
- Design document: kubernetes/community#695
- PR tracking: [kubernetes/features#368](https://github.com/kubernetes/features/issues/368#issuecomment-321625420)
Special notes for your reviewer:
First proposal submitted to the community repo, please advise if something's not right with the format or procedure, etc.
@vishh @derekwaynecarr
**Release note:**
```
NONE
```
Whenever pod sandbox needs to be recreated, all containers associated
with it will be killed by kubelet. This change ensures that the init
containers will be rerun in such cases.
The change also refactors the compute logic so that the control flow of
init containers act is more aligned with the regular containers. Unit
tests are added to verify the logic.
Automatic merge from submit-queue (batch tested with PRs 49488, 50407, 46105, 50456, 50258)
Add UpdateContainerResources method to CRI
This is first step toward support for opinionated cpu pinning for certain guaranteed pods.
In order to do this, the kubelet needs to be able to dynamically update the cpuset at the container level, which is managed by the container runtime. Thus the kubelet needs a method to communicate over the CRI so the runtime can then modify the container cgroup.
This is used in the situation where a core is added or removed from the shared pool to become a exclusive core for a new G pod. The cpuset for all containers in the shared pool will need to be updated to add or remove that core.
Opening this up now so we can start discussion. The need for a change to the CRI might be unexpected.
@derekwaynecarr @vishh @ConnorDoyle
```release-note
NONE
```
Pengfei Ni