Automatic merge from submit-queue
dockershim: Fix the race condition in ListPodSandbox
In ListPodSandbox(), we
1. List all sandbox docker containers
2. List all sandbox checkpoints. If the checkpoint does not have a
corresponding container in (1), we return partial result based on
the checkpoint.
The problem is that new PodSandboxes can be created between step (1) and
(2). In those cases, we will see the checkpoints, but not the sandbox
containers. This leads to strange behavior because the partial result
from the checkpoint does not include some critical information. For
example, the creation timestamp'd be zero, and that would cause kubelet's
garbage collector to immediately remove the sandbox.
This change fixes that by getting the list of checkpoints before listing
all the containers (since in RunPodSandbox we create them in the reverse
order).
Automatic merge from submit-queue (batch tested with PRs 42211, 38691, 42737, 42757, 42754)
Only create the symlink when container log path exists
When using `syslog` logging driver instead of `json-file`, there will not be container log files such as `<containerID-json.log>`. We should not create symlink in this case.
In ListPodSandbox(), we
1. List all sandbox docker containers
2. List all sandbox checkpoints. If the checkpoint does not have a
corresponding container in (1), we return partial result based on
the checkpoint.
The problem is that new PodSandboxes can be created between step (1) and
(2). In those cases, we will see the checkpoints, but not the sandbox
containers. This leads to strange behavior because the partial result
from the checkpoint does not include some critical information. For
example, the creation timestamp'd be zero, and that would cause kubelet's
garbage collector to immediately remove the sandbox.
This change fixes that by getting the list of checkpoints before listing
all the containers (since in RunPodSandbox we create them in the reverse
order).
1. Validate that templateGeneration is increased when and only when template is changed
2. Validate that templateGeneration is never decreased
3. Added validation tests for templateGeneration
4. Fix a bunch of errors in validate tests, for example, all validation test error cases failed
on lack of resource version, or on name changes, not on the real validation we wanted to test
Automatic merge from submit-queue
Use namespace from context
Fixes#42653
Updates rbac_test.go to submit objects without namespaces set, which matches how actual objects are submitted to the API.
Automatic merge from submit-queue (batch tested with PRs 42692, 42169, 42173)
DaemonSet: Respect ControllerRef
**What this PR does / why we need it**:
This is part of the completion of the [ControllerRef](https://github.com/kubernetes/community/blob/master/contributors/design-proposals/controller-ref.md) proposal. It brings DaemonSet into full compliance with ControllerRef. See the individual commit messages for details.
**Which issue this PR fixes**:
This ensures that DaemonSet does not fight with other controllers over control of Pods.
**Special notes for your reviewer**:
**Release note**:
```release-note
DaemonSet now respects ControllerRef to avoid fighting over Pods.
```
cc @erictune @kubernetes/sig-apps-pr-reviews
Automatic merge from submit-queue (batch tested with PRs 42692, 42169, 42173)
Add pprof trace support
Add support for `/debug/pprof/trace`
Can wait for master to reopen for 1.7.
cc @smarterclayton @wojtek-t @gmarek @timothysc @jeremyeder @kubernetes/sig-scalability-pr-reviews
The design of DaemonSet requires a relist before each phase (manage,
update, status) because it does not short-circuit and requeue for each
action triggered.
The DaemonSet Listers still use selectors, because this is the
behavior expected by callers. This clarifies the meaning of the
returned list. Some callers may need to switch to using
GetControllerOf() instead, but that is a separate, case-by-case issue.
We don't need the svcPortToInfoMap. Its only purpose was to
send "valid" local endpoints (those with valid IP and >0 port) to the
health checker. But we shouldn't be sending invalid endpoints to
the health checker anyway, because it can't do anything with them.
If we exclude invalid endpoints earlier, then we don't need
flattenValidEndpoints().
And if we don't need flattenValidEndpoints() it makes no sense to have
svcPortToInfoMap store hostPortInfo, since endpointsInfo is the same
thing as hostPortInfo except with a combined host:port.
And if svcPortToInfoMap now only stores valid endpointsInfos, it is
exactly the same thing as newEndpoints.
* Properly return ImageNotFoundError
* Support inject "Images" or "ImageInspects" and keep both in sync.
* Remove the FakeDockerPuller and let FakeDockerClient subsumes its
functinality. This reduces the overhead to maintain both objects.
* Various small fixes and refactoring of the testing utils.
Automatic merge from submit-queue (batch tested with PRs 41890, 42593, 42633, 42626, 42609)
Pods pending due to insufficient OIR should get scheduled once sufficient OIR becomes available (e2e disabled).
#41870 was reverted because it introduced an e2e test flake. This is the same code with the e2e for OIR disabled again.
We can attempt to enable the e2e test cases one-by-one in follow-up PRs, but it would be preferable to get the main fix merged in time for 1.6 since OIR is broken on master (see #41861).
cc @timothysc
Automatic merge from submit-queue (batch tested with PRs 41890, 42593, 42633, 42626, 42609)
Remove everything that is not new from batch/v2alpha1
Fixes#37166.
@lavalamp you've asked for it
@erictune this is a prereq for moving CronJobs to beta. I initially planned to put all in one PR, but after I did that I figured out it'll be easier to review separately. ptal
@kubernetes/api-approvers @kubernetes/sig-api-machinery-pr-reviews ptal
Automatic merge from submit-queue (batch tested with PRs 42506, 42585, 42596, 42584)
RC/RS: Fix ignoring inactive Pods.
**What this PR does / why we need it**:
Fix typo that broke ignoring of inactive Pods in RC, and add unit test for that case.
**Which issue this PR fixes**:
Fixes#37479
**Special notes for your reviewer**:
**Release note**:
```release-note
```
Automatic merge from submit-queue (batch tested with PRs 42506, 42585, 42596, 42584)
provide active pods to cgroup cleanup
**What this PR does / why we need it**:
This PR provides more information for when a pod cgroup is considered orphaned. The running pods cache is based on the runtime's view of the world. we create pod cgroups before containers so we should just be looking at activePods.
**Which issue this PR fixes**
Fixes https://github.com/kubernetes/kubernetes/issues/42431
Automatic merge from submit-queue
cgroup names created by kubelet should be lowercased
**What this PR does / why we need it**:
This PR modifies the kubelet to create cgroupfs names that are lowercased. This better aligns us with the naming convention for cgroups v2 and other cgroup managers in ecosystem (docker, systemd, etc.)
See: https://www.kernel.org/doc/Documentation/cgroup-v2.txt
"2-6-2. Avoid Name Collisions"
**Special notes for your reviewer**:
none
**Release note**:
```release-note
kubelet created cgroups follow lowercase naming conventions
```
Automatic merge from submit-queue (batch tested with PRs 42080, 41653, 42598, 42555)
Revert "Pods pending due to insufficient OIR should get scheduled once sufficient OIR becomes available."
Reverts kubernetes/kubernetes#41870 for stopping bleeding edge: #42597
cc/ @ConnorDoyle @kubernetes/release-team
Connor if there is a pending pr to fix the issue, please point it out to me. We can close this one, otherwise, I would like to revert the pr first. You can resubmit the fix. Thanks!
Automatic merge from submit-queue (batch tested with PRs 42080, 41653, 42598, 42555)
StatefulSet: Respect ControllerRef
**What this PR does / why we need it**:
This is part of the completion of the [ControllerRef](https://github.com/kubernetes/community/blob/master/contributors/design-proposals/controller-ref.md) proposal. It brings StatefulSet into full compliance with ControllerRef. See the individual commit messages for details.
**Which issue this PR fixes**:
Fixes#36859
**Special notes for your reviewer**:
**Release note**:
```release-note
StatefulSet now respects ControllerRef to avoid fighting over Pods. At the time of upgrade, **you must not have StatefulSets with selectors that overlap** with any other controllers (such as ReplicaSets), or else [ownership of Pods may change](https://github.com/kubernetes/community/blob/master/contributors/design-proposals/controller-ref.md#upgrading).
```
cc @erictune @kubernetes/sig-apps-pr-reviews
Deployment should ignore Pods that don't match the selector, even if
they have a ControllerRef pointing to one of the ReplicaSets it owns.
The ReplicaSet itself will orphan the Pod as soon as it syncs.
The Deployment Listers still use selectors, because this is the
behavior expected by callers. This clarifies the meaning of the
returned list. Some callers may need to switch to using
GetControllerOf() instead, but that is a separate, case-by-case issue.
The list functions in deployment/util are used outside the Deployment
controller itself. Therefore, they don't do actual adoption/orphaning.
However, they still need to avoid listing things that don't belong.
Although Deployment already applied its ControllerRef to adopt matching
ReplicaSets, it actually still used label selectors to list objects that
it controls. That meant it didn't actually follow the rules of
ControllerRef, so it could still fight with other controller types.
This should mean that the special handling for overlapping Deployments
is no longer necessary, since each Deployment will only see objects that
it owns (via ControllerRef).
Automatic merge from submit-queue
add "external" cloud provider
@thockin @justinsb @luxas @joonas
This is the discussion we had in our last meeting about moving the cloud provider dependent parts out of kubelet
1. Flex Volume v2 will not be done anytime soon, until then it is okay to leave the volume code as it is. i.e. Volume plugins with cloudprovider dependencies will continue to reside in the core kubernetes repository, and new volume plugins will be allowed.
2. We are moving forward with the previous decision to move the cloudprovider dependencies from kubelet into a new controller loop in the cloud-controller-manager.
3. The kubelet will set a taint using the flag (--register-with-taints, https://github.com/kubernetes/kubernetes/blob/master/cmd/kubelet/app/options/options.go#L238) when it starts. @luxas, we decided to go with a different approach last week, but it turns out that the other approach will not work. There is a hold on PRs that set taints in the admission controller (kubernetes/kubernetes: Pull Request 40288). Due to the uncertain nature of this hold, the kubelet flag is our only alternative. Also, since this is an alpha feature that is not fully baked, kubeadm has a lot of time to make the changes to support this, i.e. it need not support it right away. It can support it in the future releases of k8s.
4. A new flag will be added in kubelet to explicitly denote that no cloudprovider initialization should be done (--no-cloud-init). This is required to clearly indicate to the user that kubelet will not perform any cloudprovider specific calls to initialize the node object.
5. A new flag will be added in kubelet to set uniquely identifying information about the host (--external-id?). This will be required for clouds like openstack, where uniquely identifying information is only available from the host itself.
6. The upgrade scenario with these changes is straightforward, as we all are working under the assumption that the user upgrades the master before upgrading the kubelet.
In this PR, I have added the `--do-cloud-init` flag. (Item 4)
Automatic merge from submit-queue (batch tested with PRs 31783, 41988, 42535, 42572, 41870)
Pods pending due to insufficient OIR should get scheduled once sufficient OIR becomes available.
This appears to be a regression since v1.5.0 in scheduler behavior for opaque integer resources, reported in https://github.com/kubernetes/kubernetes/issues/41861.
- [X] Add failing e2e test to trigger the regression
- [x] Restore previous behavior (pods pending due to insufficient OIR get scheduled once sufficient OIR becomes available.)
Automatic merge from submit-queue (batch tested with PRs 31783, 41988, 42535, 42572, 41870)
update names for kube plugin initializer to avoid conflicts
Fixes#42581
Other API servers are likely to create admission plugin initializers and so the names we choose for our interfaces matter (they may want to run multiple initializers in the chain). This updates the names for the plugin initializers to be more specific. No other changes.
@ncdc
Automatic merge from submit-queue (batch tested with PRs 31783, 41988, 42535, 42572, 41870)
controller: ensure deployment rollback is re-entrant
Make rollbacks re-entrant in the Deployment controller, otherwise
fast enqueues of a Deployment may end up in undesired behavior
- redundant rollbacks.
Fixes https://github.com/kubernetes/kubernetes/issues/36703
@kubernetes/sig-apps-bugs
The StatefulSet Listers still use selectors, because this is the
behavior expected by callers. This clarifies the meaning of the
returned list. Some callers may need to switch to using
GetControllerOf() instead, but that is a separate, case-by-case issue.
Automatic merge from submit-queue
make the system:authenticated group adder smarter
Fixes#42437
This prevents the group adder from adding the system:authenticated group when:
1. it's already in the list
2. the user is system:anonymous
3. system:unauthenticated is in the list
Smaller alternative to https://github.com/kubernetes/kubernetes/pull/42421 for 1.6.
@kubernetes/sig-auth-pr-reviews @enj @liggitt
Make rollbacks re-entrant in the Deployment controller, otherwise
fast enqueues of a Deployment may end up in undesired behavior
- redundant rollbacks.
Automatic merge from submit-queue
Fix Multizone pv creation on GCE
When Multizone is enabled static PV creation on GCE
fails because Cloud provider configuration is not
available in admission plugins.
cc @derekwaynecarr @childsb
Automatic merge from submit-queue
[Bug] Fix gpu initialization in Kubelet
Kubelet incorrectly fails if `AllAlpha=true` feature gate is enabled with container runtimes that are not `docker`.
Replaces #42407
Automatic merge from submit-queue
kubectl: respect deployment strategy parameters for rollout status
Fixes https://github.com/kubernetes/kubernetes/issues/40496
`rollout status` now respects the strategy parameters for a RollingUpdate Deployment. This means that it will exit as soon as minimum availability is reached for a rollout (note that if you allow maximum availability, `rollout status` will succeed as soon as the new pods are created)
@janetkuo @AdoHe ptal
Automatic merge from submit-queue
Small fix to the bootstrap TokenCleaner
Accidentally missed setting options and so the TokenCleaner was in a retry loop. Also moved from using an explicit timer over cached values vs. relying on a short resync timeout.
```release-note
```
Putting this in the 1.6 milestone as this is clearly a bug fix in a new feature.
- Added schedulercache.Resource.SetOpaque helper.
- Amend kubelet allocatable sync so that when OIRs are removed from capacity
they are also removed from allocatable.
- Fixes#41861.
Automatic merge from submit-queue (batch tested with PRs 42456, 42457, 42414, 42480, 42370)
In DaemonSet e2e test, don't check nodes with NoSchedule taints
Fixes#42345
For example, master node has a ismaster:NoSchedule taint. We don't expect pods to be created there without toleration.
cc @marun @lukaszo @kargakis @yujuhong @Random-Liu @davidopp @kubernetes/sig-apps-pr-reviews
Automatic merge from submit-queue (batch tested with PRs 42456, 42457, 42414, 42480, 42370)
controller: reduce log verbosity for deployments
Fixes https://github.com/kubernetes/kubernetes/issues/41187
Labeling as a bug fix since I think excessive logging should be considered as a bug.
@kubernetes/sig-apps-bugs
Automatic merge from submit-queue (batch tested with PRs 42369, 42375, 42397, 42435, 42455)
[Bug Fix]: Avoid evicting more pods than necessary by adding Timestamps for fsstats and ignoring stale stats
Continuation of #33121. Credit for most of this goes to @sjenning. I added volume fs timestamps.
**why is this a bug**
This PR attempts to fix part of https://github.com/kubernetes/kubernetes/issues/31362 which results in multiple pods getting evicted unnecessarily whenever the node runs into resource pressure. This PR reduces the chances of such disruptions by avoiding reacting to old/stale metrics.
Without this PR, kubernetes nodes under resource pressure will cause unnecessary disruptions to user workloads.
This PR will also help deflake a node e2e test suite.
The eviction manager currently avoids evicting pods if metrics are old. However, timestamp data is not available for filesystem data, and this causes lots of extra evictions.
See the [inode eviction test flakes](https://k8s-testgrid.appspot.com/google-node#kubelet-flaky-gce-e2e) for examples.
This should probably be treated as a bugfix, as it should help mitigate extra evictions.
cc: @kubernetes/sig-storage-pr-reviews @kubernetes/sig-node-pr-reviews @vishh @derekwaynecarr @sjenning
Automatic merge from submit-queue (batch tested with PRs 42369, 42375, 42397, 42435, 42455)
Kubelet: return container runtime's version instead of CRI's one
**What this PR does / why we need it**:
With CRI enabled by default, kubelet reports the version of CRI instead of container runtime version. This PR fixes this problem.
**Which issue this PR fixes**
Fixes#42396.
**Special notes for your reviewer**:
Should also cherry-pick to 1.6 branch.
**Release note**:
```release-note
NONE
```
cc @yujuhong @kubernetes/sig-node-bugs
Automatic merge from submit-queue
Output of `kubectl get` is inconsistent for pods
Builds on top of fixes from #42283, only the last two commits are new. Reverts behavior of #39042 which was inconsistent and confusing.
Fixes#15853
Accidentally missed setting options and so the TokenCleaner was in a retry loop. Also moved from using an explicit timer over cached values vs. relying on a short resync timeout.
Signed-off-by: Joe Beda <joe.github@bedafamily.com>
Automatic merge from submit-queue
Eviction Manager Enforces Allocatable Thresholds
This PR modifies the eviction manager to enforce node allocatable thresholds for memory as described in kubernetes/community#348.
This PR should be merged after #41234.
cc @kubernetes/sig-node-pr-reviews @kubernetes/sig-node-feature-requests @vishh
** Why is this a bug/regression**
Kubelet uses `oom_score_adj` to enforce QoS policies. But the `oom_score_adj` is based on overall memory requested, which means that a Burstable pod that requested a lot of memory can lead to OOM kills for Guaranteed pods, which violates QoS. Even worse, we have observed system daemons like kubelet or kube-proxy being killed by the OOM killer.
Without this PR, v1.6 will have node stability issues and regressions in an existing GA feature `out of Resource` handling.
Automatic merge from submit-queue (batch tested with PRs 42443, 38924, 42367, 42391, 42310)
Apply custom defaults to init containers
Adds overridden defaults to init containers. They were not being defaulted the same way normal containers were.
Automatic merge from submit-queue (batch tested with PRs 42443, 38924, 42367, 42391, 42310)
Dell EMC ScaleIO Volume Plugin
**What this PR does / why we need it**
This PR implements the Kubernetes volume plugin to allow pods to seamlessly access and use data stored on ScaleIO volumes. [ScaleIO](https://www.emc.com/storage/scaleio/index.htm) is a software-based storage platform that creates a pool of distributed block storage using locally attached disks on every server. The code for this PR supports persistent volumes using PVs, PVCs, and dynamic provisioning.
You can find examples of how to use and configure the ScaleIO Kubernetes volume plugin in [examples/volumes/scaleio/README.md](examples/volumes/scaleio/README.md).
**Special notes for your reviewer**:
To facilitate code review, commits for source code implementation are separated from other artifacts such as generated, docs, and vendored sources.
```release-note
ScaleIO Kubernetes Volume Plugin added enabling pods to seamlessly access and use data stored on ScaleIO volumes.
```
Automatic merge from submit-queue (batch tested with PRs 41919, 41149, 42350, 42351, 42285)
enable cgroups tiers and node allocatable enforcement on pods by default.
```release-note
Pods are launched in a separate cgroup hierarchy than system services.
```
Depends on #41753
cc @derekwaynecarr
Automatic merge from submit-queue (batch tested with PRs 41919, 41149, 42350, 42351, 42285)
kubelet: enable qos-level memory limits
```release-note
Experimental support to reserve a pod's memory request from being utilized by pods in lower QoS tiers.
```
Enables the QoS-level memory cgroup limits described in https://github.com/kubernetes/community/pull/314
**Note: QoS level cgroups have to be enabled for any of this to take effect.**
Adds a new `--experimental-qos-reserved` flag that can be used to set the percentage of a resource to be reserved at the QoS level for pod resource requests.
For example, `--experimental-qos-reserved="memory=50%`, means that if a Guaranteed pod sets a memory request of 2Gi, the Burstable and BestEffort QoS memory cgroups will have their `memory.limit_in_bytes` set to `NodeAllocatable - (2Gi*50%)` to reserve 50% of the guaranteed pod's request from being used by the lower QoS tiers.
If a Burstable pod sets a request, its reserve will be deducted from the BestEffort memory limit.
The result is that:
- Guaranteed limit matches root cgroup at is not set by this code
- Burstable limit is `NodeAllocatable - Guaranteed reserve`
- BestEffort limit is `NodeAllocatable - Guaranteed reserve - Burstable reserve`
The only resource currently supported is `memory`; however, the code is generic enough that other resources can be added in the future.
@derekwaynecarr @vishh
When gce cloud tries to delete a disk, if the disk could not be found
from the zones, the function should return nil error. This modified behavior is also consistent with AWS
Automatic merge from submit-queue (batch tested with PRs 41306, 42187, 41666, 42275, 42266)
Implement bulk polling of volumes
This implements Bulk volume polling using ideas presented by
justin in https://github.com/kubernetes/kubernetes/pull/39564
But it changes the implementation to use an interface
and doesn't affect other implementations.
cc @justinsb
Automatic merge from submit-queue (batch tested with PRs 42365, 42429, 41770, 42018, 35055)
controller: statefulsets respect observed generation
StatefulSets do not update ObservedGeneration even though the API field is in place. This means that clients can never be sure whether the StatefulSet controller has observed the latest spec of a StatefulSet.
@kubernetes/sig-apps-bugs
This implements Bulk volume polling using ideas presented by
justin in https://github.com/kubernetes/kubernetes/pull/39564
But it changes the implementation to use an interface
and doesn't affect other implementations.
Automatic merge from submit-queue (batch tested with PRs 41984, 41682, 41924, 41928)
make edit respect --save-config
New behavior:
edit:
1) the flag is set to true: it will create or update the last-applied-config annotation
2) the flag is set to false or is unspecified: NOP
replace:
1) If saveConfig is true, create or update the annotation.
2) If saveConfig is false and the local config file doesn't have the annotation, we save the annotation from the live object if there is one (Try to not surprise the users).
3) If saveConfig is false and the local config file has the annotation, we use the annotation in the config file.
4) Same behavior for force replacing
fixes#40626
```release-note
stop kubectl edit from updating the last-applied-configuration annotation when --save-config is unspecified or false.
```
@pwittrock @liggitt
This is a bug fix that prevented `edit` from being respected by `apply`
Automatic merge from submit-queue (batch tested with PRs 41984, 41682, 41924, 41928)
RC/RS: Fully Respect ControllerRef
**What this PR does / why we need it**:
This is part of the completion of the [ControllerRef](https://github.com/kubernetes/community/blob/master/contributors/design-proposals/controller-ref.md) proposal. It brings ReplicaSet and ReplicationController into full compliance with ControllerRef. See the individual commit messages for details.
**Which issue this PR fixes**:
Although RC/RS had partially implemented ControllerRef, they didn't use it to determine which controller to sync, or to update expectations. This could lead to instability or controllers getting stuck.
Ref: https://github.com/kubernetes/kubernetes/issues/24433
**Special notes for your reviewer**:
**Release note**:
```release-note
```
cc @erictune @kubernetes/sig-apps-pr-reviews
Automatic merge from submit-queue (batch tested with PRs 42128, 42064, 42253, 42309, 42322)
Add storage.k8s.io/v1 API
This is combined version of reverted #40088 (first 4 commits) and #41646. The difference is that all controllers and tests use old `storage.k8s.io/v1beta1` API so in theory all tests can pass on GKE.
Release note:
```release-note
StorageClassName attribute has been added to PersistentVolume and PersistentVolumeClaim objects and should be used instead of annotation `volume.beta.kubernetes.io/storage-class`. The beta annotation is still working in this release, however it will be removed in a future release.
```
Automatic merge from submit-queue (batch tested with PRs 41672, 42084, 42233, 42165, 42273)
ExecProbes should be able to do simple env var substitution
For containers that don't have bash, we should support env substitution
like we do on command and args. However, without major refactoring
valueFrom is not supportable from inside the prober. For now, implement
substitution based on hardcoded env and leave TODOs for future work.
Improves the state of #40846, will spawn a follow up issue for future refactoring after CRI settles down
Automatic merge from submit-queue (batch tested with PRs 41672, 42084, 42233, 42165, 42273)
Don't sync IPtables before underlying store/reflector is fully synced
Ref #42108
Build on top of #42108 - only the second commit is unique.
Automatic merge from submit-queue (batch tested with PRs 42126, 42130, 42232, 42245, 41932)
fix LeaderTransitions always zero
on leader transition, LeaderTransitions is increased to 1, but
then cleared to zero by next renew.
External monitoring system may watch LeaderTransitions and rely
on it's facticity.
Kubernetes Submit Queue