optionsExternalVersion is being used for shared types (meta.k8s.io). The
installer should first look in the current API group for GET and CONNECT
options objects before checking in `v1`.
Kubernetes-commit: 0e4b20fbf9ff299c37a5055e082fa27dbf2b49c0
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
```
Kubernetes-commit: 131b419596a3d4fce1020c3b11f78034b5880593
The etcd3 storage now attempts to fill partial pages to prevent clients
having to make more round trips (latency from server to etcd is lower
than client to server). The server makes repeated requests to etcd of
the current page size, then uses the filter function to eliminate any
matches. After this change the apiserver will always return full pages,
but we leave the language in place that clients must tolerate it.
Reduces tail latency of large filtered lists, such as viewing pods
assigned to a node.
Kubernetes-commit: da7124e5e5c0385dd5bcfc72ef035effc7708913
Automatic merge from submit-queue (batch tested with PRs 52843, 52710, 52821, 52844). 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>..
update-staging-godeps: only mangle staging repos in staging Godeps.json
- this re-adds non-staging k8s.io/* repos to the staging Godeps.jsons
- x-outs instead of removing of staging dependencies in order to get a
precise trigger for a complete godep restore+save run in the publisher
bot.
The first breaks k8s.io/kube-aggregator's staging export.
The second potentially leads to inconsistent godeps in our exported staging repos.
Kubernetes-commit: 3ba46ee9fab32ccaeeee4af4a80375e6e1d71ace
Automatic merge from submit-queue (batch tested with PRs 52477, 52790, 52798). 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>..
conversion-gen: make staging dirs independent of living in vendor/
The `+k8s:conversion-gen` tags included the package directory inside of kube's vendor dir. This
makes them invalid when we publish staging repos.
Without this PR our sample-apiserver example code-generation is broken once published.
Kubernetes-commit: 74ff323f9e546e62288d331f62a37e0ec58f6b35
Currently openstack cloud provider just support keystone v2.0 and v3
The latest Identity Service is publishing an ID of v3.8, we should
update gophercloud to recognize v3.8 as a valid version id.
Kubernetes-commit: f98326b7f9364f6a5c3e336af1279833713e73c9
Allows a caller to know how many exec, log, proxy, and watch calls are
running at the current moment.
Kubernetes-commit: fabce1b893f96bdf466c1fdb1fcf825210c008ae
Automatic merge from submit-queue (batch tested with PRs 52176, 43152). 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>..
etcd3 store: retry with live object on conflict if there was a suggestion
Retry with a live object instead of the cached version if the watch
cache receives a conflict trying to do the update.
Fixes#41892
Kubernetes-commit: d48611a1da00a2fe32011404fe25fb6af51673a5
In GuaranteedUpdate, if it was called with a suggestion (e.g. via the
watch cache), and the suggested object is stale, perform a live lookup
and then retry the update.
Signed-off-by: Andy Goldstein <andy.goldstein@gmail.com>
Kubernetes-commit: bf33df16b52508974ddedacd814010cfe0fb79f0
Automatic merge from submit-queue
Increase sliding window to 5hr for request_latencies metric
We're seeing high latency values for a couple of types of api calls in our density test (ref https://github.com/kubernetes/kubernetes/issues/51899). And we're recording values from only the last 1 hour in the metric (as @wojtek-t told me offline) - so our test result is pretty much counting only the calls during the delete phase.
cc @kubernetes/sig-scalability-misc @kubernetes/sig-api-machinery-misc @gmarek
Kubernetes-commit: 9ef9a1b8f032ecccee94aff773e9db13b9b7ced6
Automatic merge from submit-queue
Add extra steps to delete resource handler trace
Based on https://github.com/kubernetes/kubernetes/issues/51899#issuecomment-329786131
cc @kubernetes/sig-scalability-misc @kubernetes/sig-api-machinery-misc @gmarek
Kubernetes-commit: 9f954c146f37261aa62981b7505f9ba0dcb255e1
Automatic merge from submit-queue (batch tested with PRs 51824, 50476, 52451, 52009, 52237)
Improve apiserver metrics reporting
Normalize "WATCHLIST" to "WATCH", add "scope" to the other metrics (listing 50k pods is != listing pods in a namespace), and add a new scope "resource" to cover individual resource calls.
This roughly aligns metrics with our ACL model (technically resource scope is GET, but POST to a subresource and POST to a namespace are not the same thing).
```release-note
WATCHLIST calls are now reported as WATCH verbs in prometheus for the apiserver_request_* series. A new "scope" label is added to all apiserver_request_* values that is either 'cluster', 'resource', or 'namespace' depending on which level the query is performed at.
```
Kubernetes-commit: 93ddb7be5f547f5ebb0a60bd1135234fdf4e3c07
Remove the MonitorRequest method and replace with a method that takes
request.RequestInfo, which is our default way to talk about API objects.
Preserves existing semantics for calls.
Kubernetes-commit: 10e6dc5ed3573118c56fa8823b387e47c8e8ae06
Automatic merge from submit-queue (batch tested with PRs 50949, 52155, 52175, 52112, 52188)
Allow watch cache to be disabled per type
Currently setting watch cache size for a given resource does not disable
the watch cache. This commit adds a new `default-watch-cache-size` flag
to map to the existing field, and refactors how watch cache sizes are
calculated to bring all of the code into one place. It also adds debug
logging to startup to allow us to verify watch cache enablement in
production.
Part of #51825
Will allow watch cache to be disabled selectively.
Kubernetes-commit: e1bf145c9fb8766ee3be3f0e06175dd9da3cbe5d
Currently setting watch cache size for a given resource does not disable
the watch cache. This commit adds a new `default-watch-cache-size` flag
to map to the existing field, and refactors how watch cache sizes are
calculated to bring all of the code into one place. It also adds debug
logging to startup to allow us to verify watch cache enablement in
production.
Kubernetes-commit: fc2d201e155296f311ae0a9278b00dcae2d68708
Trying to make sure we always log the flags an instance of apiserver
starts with.
This can be especially valuable for emailed logs or e2e/kubemark tests.
Kubernetes-commit: 366459aee84a4de1f0342a464bd479a5d12e6f0f
Automatic merge from submit-queue (batch tested with PRs 51900, 51782, 52030)
Fill in creationtimestamp in audit events
**What this PR does / why we need it**:
This is fixing null creationtimestamp in audit events.
@sttts @crassirostris like we've talked earlier today
**Release note**:
```release-note
none
```
Kubernetes-commit: 36b3a0d75b177f7fb3b68750b46dd4c2aeb45664
Automatic merge from submit-queue (batch tested with PRs 51900, 51782, 52030)
A policy with 0 rules should return an error
**Which issue this PR fixes**
[isuue#51565](https://github.com/kubernetes/kubernetes/issues/51565)
**Release note**:
```
An audit policy file with 0 rule returns an error.
```
Kubernetes-commit: 4a72b32d3693e4f53c004c48d7ff58f628289fa6
Automatic merge from submit-queue (batch tested with PRs 51900, 51782, 52030)
apiservers: stratify versioned informer construction
The versioned share informer factory has been part of the GenericApiServer config,
but its construction depended on other fields of that config (e.g. the loopback
client config). Hence, the order of changes to the config mattered.
This PR stratifies this by moving the SharedInformerFactory from the generic Config
to the CompleteConfig struct. Hence, it is only filled during completion when it is
guaranteed that the loopback client config is set.
While doing this, the CompletedConfig construction is made more type-safe again,
i.e. the use of SkipCompletion() is considereably reduced. This is archieved by
splitting the derived apiserver Configs into the GenericConfig and the ExtraConfig
part. Then the completion is structural again because CompleteConfig is again
of the same structure: generic CompletedConfig and local completed ExtraConfig.
Fixes#50661.
Kubernetes-commit: 63d6bdb58c864de4f42ff139cb0a044a35f7a36f
Automatic merge from submit-queue (batch tested with PRs 52091, 52071)
Log a warning when --audit-policy-file not passed to apiserver
When audit backend provided, but --audit-policy-file is not passed, leave a info in log.
**Release note**:
```
Log a warning when --audit-policy-file not passed to apiserver
```
Kubernetes-commit: 500642602210088447c9dcd7abe2e26815f271ed
Automatic merge from submit-queue
fix format of forbidden messages
**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#51813
**Special notes for your reviewer**:
/assign @deads2k @liggitt
**Release note**:
```release-note
None
```
Kubernetes-commit: 45fe0a9e0417eb148c3082c53ae16da7377a2c32
Clayton Coleman