Remove the "OrDie" from the name (since it doesn't "or die") and add
an extra check that there is at least 1 node available, since many
callers already did that themselves, and many others should have.
Errors from staticcheck:
cmd/kube-scheduler/app/server.go:297:27: prometheus.Handler is deprecated: Please note the issues described in the doc comment of InstrumentHandler. You might want to consider using promhttp.Handler instead. (SA1019)
pkg/apis/scheduling/v1alpha1/defaults.go:27:6: func addDefaultingFuncs is unused (U1000)
pkg/apis/scheduling/v1beta1/defaults.go:27:6: func addDefaultingFuncs is unused (U1000)
test/e2e/scheduling/predicates.go:757:6: func verifyReplicasResult is unused (U1000)
test/e2e/scheduling/predicates.go:765:6: func getPodsByLabels is unused (U1000)
test/e2e/scheduling/predicates.go:772:6: func runAndKeepPodWithLabelAndGetNodeName is unused (U1000)
test/e2e/scheduling/limit_range.go:172:3: this value of pod is never used (SA4006)
test/e2e/scheduling/limit_range.go:177:3: this value of pod is never used (SA4006)
test/e2e/scheduling/limit_range.go:196:3: this value of pod is never used (SA4006)
test/e2e/scheduling/limit_range.go:201:3: this value of pod is never used (SA4006)
test/e2e/scheduling/limit_range.go:240:3: this value of pod is never used (SA4006)
test/e2e/scheduling/taints.go:428:13: this value of err is never used (SA4006)
test/e2e/scheduling/ubernetes_lite.go:219:2: this value of pods is never used (SA4006)
test/integration/scheduler/extender_test.go:78:4: this value of resp is never used (SA4006)
test/integration/volumescheduling/volume_binding_test.go:529:15: this result of append is never used, except maybe in other appends (SA4010)
test/integration/volumescheduling/volume_binding_test.go:538:15: this result of append is never used, except maybe in other appends (SA4010)
Originally, printOnce() was implemented for printing debugging message
once for checking actual value is the same as expected value.
However even if an error doesn't happen in ExpectEqual(), printOnce()
is called internally. Then the second ExpectEqual() doesn't output
the debugging message if an error happens. This makes debugging
difficult, let's just remove it for debugging.
The conformance test "validates resource limits of pods that are
allowed to run" is failed if master node doesn't have taints.
The reason is that nodeList which should consist of schedulable
nodes is gotten with GetMasterAndWorkerNodes() and master nodes
were excluded from nodeList.
This makes nodeList be gotten with GetReadySchedulableNodesOrDie()
and nodeList can consist of schedulable nodes.
"validates resource limits of pods that are allowed to run" test of
conformance tests are flake on some local environments.
The CPU workload pods don't seem work well and nodes have still CPU
capacity after running the workload pods. Then the conformance test
failed unexpectedly.
This adds message which shows how much CPU used by the workload pods
for investigating it easily.
Automatic merge from submit-queue (batch tested with PRs 61546, 61038, 61575, 60779, 61496). 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>.
Use random port to avoid conflict with other important pods that might be listening on 80.
**What this PR does / why we need it**:
Fix https://github.com/kubernetes/kubernetes/issues/61018
**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/61018
**Special notes for your reviewer**:
**Release note**:
```release-note
None.
```
@smarterclayton
This commit fixes an issue where in clusters which have FQDN as the node names,
one of the scheduling predicates tests will fail because it will try and run a
pod with a name that violates DNS-1123 rules. As an example, one such pod name
could look like "filler-pod-kube-node-0.kubelet.mesos".
Signed-off-by: Paulo Pires <pjpires@gmail.com>
Automatic merge from submit-queue
Version gates the ephemeral storage e2e test
Version gates the ephemeral storage e2e test.
**Release note**:
```
NONE
```
@kubernetes/sig-testing-pr-reviews
Automatic merge from submit-queue
Add pod preemption to the scheduler
**What this PR does / why we need it**:
This is the last of a series of PRs to add priority-based preemption to the scheduler. This PR connects the preemption logic to the scheduler workflow.
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: fixes#48646
**Special notes for your reviewer**:
This PR includes other PRs which are under review (#50805, #50405, #50190). All the new code is located in 43627afdf9.
**Release note**:
```release-note
Add priority-based preemption to the scheduler.
```
ref/ #47604
/assign @davidopp
@kubernetes/sig-scheduling-pr-reviews
Automatic merge from submit-queue (batch tested with PRs 50257, 50247, 50665, 50554, 51077)
Replace hard-code "cpu" and "memory" to consts
**What this PR does / why we need it**:
There are many places using hard coded "cpu" and "memory" as resource name. This PR replace them to consts.
**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**:
/kind cleanup
**Release note**:
```release-note
NONE
```
Automatic merge from submit-queue (batch tested with PRs 49358, 49253)
Remove hostname label condition in SchedulerPredicates
**What this PR does / why we need it**:
```
validates that NodeSelector is respected if matching [Conformance]
validates that required NodeAffinity setting is respected if matching
```
The two tests above make the assumption that the node names are equal to the `kubernetes.io/hostname` labels. Unfortunately, this is not necessarily true all the time. For instance, when using the AWS Cloud Provider + Container Linux:
- The node name is set using the AWS SDK's `ec2.Instance.PrivateDnsName` and has the form `ip-10-0-35-57.ca-central-1.compute.internal` [[1](https://github.com/kubernetes/kubernetes/blob/v1.7.1/pkg/cloudprovider/providers/aws/aws.go#L3343-L3346)] [[2](https://raw.githubusercontent.com/aws/aws-sdk-go/master/service/ec2/api.go)]
- The node's hostname, however, is a simple call to `os.Hostname()`, itself reading `/proc/sys/kernel/hostname`, which contains what the AWS DHCP assigned to the instance, typically the hostname short-form: `ip-10-0-16-137`. [[1](https://github.com/kubernetes/kubernetes/blob/v1.7.1/pkg/util/node/node.go#L43-L54)]
Consequently, we are trying to assign a pod to a node having the following label: `kubernetes.io/hostname=ip-10-0-35-57.ca-central-1.compute.internal` (in addition to the randomly generated label), whereas the actual label on the node is `kubernetes.io/hostname=ip-10-0-35-57`.
Furthermore, this inaccurate `kubernetes.io/hostname=<nodename>` condition is actually useless given we already match over a random label, that was assigned to that node. Later, the test ensures that the scheduled pod was scheduled to the right node by comparing the pod's node name and the node name we expected the pod to be on:
```
framework.ExpectNoError(framework.WaitForPodNotPending(cs, ns, labelPodName))
labelPod, err := cs.Core().Pods(ns).Get(labelPodName, metav1.GetOptions{})
framework.ExpectNoError(err)
Expect(labelPod.Spec.NodeName).To(Equal(nodeName))
```
The `k8s.io/apimachinery/pkg/types/nodename` data structure actually [warns](55bee3ad21/staging/src/k8s.io/apimachinery/pkg/types/nodename.go (L40-L43)) about the fact that the node name might be different than the hostname on AWS.
**Release note**:
```release-note
NONE
```
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
Delete unused return
**What this PR does / why we need it**:
We do not use the function return, it's better not to write the return.
**Release note**:
```release-note
NONE
```
Automatic merge from submit-queue (batch tested with PRs 45610, 47628)
Replace capacity with allocatable to calculate pod resource
It is not accurate to use capacity to do the calculation.
**What this PR does / why we need it**:
The currently cpu resource calculation for a pod in end2end test is incorrect.
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: fixes #
fixes#47627
**Special notes for your reviewer**:
More details about capacity and allocatable:
https://github.com/kubernetes/community/blob/master/contributors/design-proposals/node-allocatable.md
**Release note**:
NONE