Sok3t/descheduler
1
0
Fork 0
mirror of https://github.com/kubernetes-sigs/descheduler.git synced 2026-01-26 13:29:11 +01:00
Code Packages Releases Activity
509 Commits 21 Branches 83 Tags
8c7267b3791c82bf546b5c82a84ebba8c1008074
Commit Graph

218 Commits

Author SHA1 Message Date
Jan Chaloupka
eacbae72fd Mention pod eviction is in dry mode when set 
Indicate whether a pod is only virtually evicted so there's no confusion
with the real eviction.
 2020-05-20 10:40:58 +02:00
Mike Dame
c6ff87dbd6 Consider pod image in duplicates strategy 
This increases the specificity of the RemoveDuplicates strategy by
removing pods which not only have the same owner, but who also
must have the same list of container images. This also adds a
parameter, `ExcludeOwnerKinds` to the RemoveDuplicates strategy
which accepts a list of Kinds. If a pod has any of these Kinds as
an owner, that pod is not considered for eviction.
 2020-05-15 09:37:09 -04:00
Sean Malloy
55afde6251 Remove line break in log message from PodLifeTime strategy 2020-05-14 00:36:52 -05:00
Sean Malloy
7039b6c8aa Refactor function EvictPod to only return an error 
The EvictPod function previously returned a bool and an error. The
function now only returns an error. Callers can check for failure by
testing if the returned error is not nil. This aligns the EvictPod
function with idiomatic Go best practices.

Also, the function name has been changed from EvictPod to evictPod
because it is not used outside the evictions package.
 2020-05-14 00:11:54 -05:00
Sean Malloy
cff984261e Log an error when EvictPod method returns a non-nil error 
End users should be able to see the detailed error from the EvictPod
method when it fails. Updates all strategies to log the error. The
PodLifeTime strategy already logs this error.
 2020-05-13 23:36:32 -05:00
Sean Malloy
4819ab9c69 Add namespace to pod eviction log messages 
In multi-tenant environments it is useful to know which namespace a pod
was evicted from. Therefore log the namespace when evicting pods.

Also, do not log a nil error when successfully evicting pods.
 2020-05-13 23:28:25 -05:00
Mike Dame
eec1104d6e React to 1.18 by adding contexts to client calls 2020-05-12 15:01:25 -04:00
Sean Malloy
423ee35846 Add New PodLifeTime Strategy 
The new PodLifeTime descheduler strategy can be used to evict pods that
were created more than the configured number of seconds ago.

In the below example pods created more than 24 hours ago will be evicted.
````
apiVersion: "descheduler/v1alpha1"
kind: "DeschedulerPolicy"
strategies:
  "PodLifeTime":
     enabled: true
     params:
        maxPodLifeTimeSeconds: 86400
````
 2020-05-07 23:10:36 -05:00
Mike Dame
fa335c782f Switch PodsHavingTooManyRestarts params to pointer 2020-05-06 13:29:05 -04:00
Jan Chaloupka
cbcefb5d2f Remove options.DeschedulerServer from all strategies 2020-04-28 16:13:33 +02:00
Jan Chaloupka
991eddb691 Turn StrategyParameters.NodeResourceUtilizationThresholds field into a pointer 
The field is intended to be omitempty when not set. Without a pointer the strategy
serialized into json string looks like:

```json
strategies:
  LowNodeUtilization:
    enabled: true
    params:
      nodeResourceUtilizationThresholds:
        numberOfNodes: 1
        targetThresholds:
          cpu: 50
          memory: 50
          pods: 20
        thresholds:
          cpu: 50
          memory: 50
          pods: 20
  RemoveDuplicates:
    enabled: true
    params:
      nodeResourceUtilizationThresholds: {}
  RemovePodsViolatingInterPodAntiAffinity:
    enabled: true
    params:
      nodeResourceUtilizationThresholds: {}
  RemovePodsViolatingNodeAffinity:
    enabled: true
    params:
      nodeAffinityType:
      - requiredDuringSchedulingIgnoredDuringExecution
      nodeResourceUtilizationThresholds: {}
  RemovePodsViolatingNodeTaints:
    enabled: true
    params:
      nodeResourceUtilizationThresholds: {}
```

It's preferred to have the following json string instead:
```
strategies:
  LowNodeUtilization:
    enabled: true
    params:
      nodeResourceUtilizationThresholds:
        numberOfNodes: 1
        targetThresholds:
          cpu: 50
          memory: 50
          pods: 20
        thresholds:
          cpu: 50
          memory: 50
          pods: 20
  RemoveDuplicates:
    enabled: true
  RemovePodsViolatingInterPodAntiAffinity:
    enabled: true
  RemovePodsViolatingNodeAffinity:
    enabled: true
    params:
      nodeAffinityType:
      - requiredDuringSchedulingIgnoredDuringExecution
  RemovePodsViolatingNodeTaints:
    enabled: true
```
 2020-04-28 15:20:30 +02:00
Mike Dame
e7c42794a0 Add RemovePodsHavingTooManyRestarts strategy 2020-04-24 10:48:28 -04:00
Jan Chaloupka
077b7f6505 Drop check for MaxNoOfPodsToEvictPerNode and invoke EvictPod wrapper instead 2020-04-22 10:33:22 +02:00
Jan Chaloupka
240fa93bc5 Move maximum-pods-per-nodes-evicted logic under a single invocation 
Each strategy implements a test for checking if a maximum number of pods per node
was already evicted. The test duplicates a code that can be put
under a single invocation. Thus, reducing the number of arguments passed
to each strategy given EvicPod call can encapsulate both the check
and the invocation of the pod eviction itself.
 2020-04-22 10:32:16 +02:00
Kubernetes Prow Robot
6c7f846917 Merge pull request #265 from ingvagabund/tolerations-tains-code-cleanup 
Tolerations taints code cleanup
 2020-04-21 14:27:52 -07:00
Jan Chaloupka
6db7c3b92c Call utils.TolerationsTolerateTaintsWithFilter directly, not through checkPodsSatisfyTolerations 2020-04-21 22:34:14 +02:00
Jan Chaloupka
1c300a9881 Drop getNoScheduleTaints and allTaintsTolerated in favor of utils.TolerationsTolerateTaintsWithFilter 2020-04-19 22:18:49 +02:00
Jan Chaloupka
0e9b33b822 allTaintsTolerated: remove for iteration through tolerations which is already implemented in utils.TolerationsTolerateTaint 2020-04-19 22:07:34 +02:00
Jan Chaloupka
36e3d1e703 Drop local implementation of toleratesTaint in favor of k8s.io/api/core/v1.Toleration.TolerateTaint 
Functionally identical implementation of toleratesTaint is already provided in k8s.io/api
 2020-04-19 21:58:41 +02:00
Jan Chaloupka
f53264b613 lownodeutilization: evict best-effort pods only 
Unit test refactored node utilization and pod clasification
 2020-04-17 11:32:34 +02:00
Jan Chaloupka
414554ae5e lownodeutilization: make unit tests with/without priority table driven 2020-04-17 11:32:28 +02:00
Jan Chaloupka
150f945592 lownodeutilization: clasify pods of over utilized nodes only 
Only over utilized nodes need clasification of pods into categories.
Thus, skipping categorizing of pods which saves computation time in cases
where the number of over utilized nodes makes less than 50% of all nodes
or their fraction.
 2020-04-17 11:32:22 +02:00
Mike Dame
e95e42930d Remove early return false from PodFitsAnyNode 2020-04-02 11:53:29 -04:00
Kubernetes Prow Robot
112684bcb9 Merge pull request #181 from jw-s/feature/respect-tolerations 
Respect Node taints with tolerations (if exist)
 2020-03-04 10:45:48 -08:00
Jan Chaloupka
9593ce16d9 List nodes through informer in every iteration 
Also, refactor the code a bit so it can be tested without checking for eviction support.
 2020-03-04 16:26:13 +01:00
Joel Whittaker-Smith
4757132452 check for taints on low nodes and honor them 2020-03-02 14:43:11 +01:00
Avesh Agarwal
8dc7b475d9 Fix makefile for make gen, and Update hack files and 
auto-generated files.
 2020-02-07 16:32:55 -05:00
Kubernetes Prow Robot
fc8c581d7a Merge pull request #226 from ankon/pr/node-unschedulable-typo 
Fix typo in function name
 2020-02-07 08:13:47 -08:00
Jan Chaloupka
d065f9904b Critical pod annotation scheduler.alpha.kubernetes.io/critical-pod has been drop, no need to set it anymore in the tests 2020-02-07 16:29:44 +01:00
Andreas Kohn
c3346e9806 Fix typo in function name 2020-02-04 10:03:08 +01:00
Mike Dame
816f8cb682 Fix build errors 2020-02-03 10:05:04 -05:00
Mike Dame
431597dd43 Break dependency on k8s.io/kubernetes 2020-02-03 10:04:53 -05:00
Mike Dame
30d05382b6 Wire --descheduler-interval flag to run descheduler on a loop 2020-01-30 17:17:55 -05:00
Sean Malloy
e3d06d1541 Create k8s event when a pod is successfully evicted 
This allows end users to more easily monitor when the descheduler evicts
a pod in their environment. Prior to adding k8s event support the only
way to track evicted pods was to view the logs.
 2019-12-11 21:35:57 -06:00
Zhang Jinghui
872953b9cf normalize blank import 2019-11-09 21:53:29 +08:00
swatisehgal
7563b5561b Strategy to consider taints and tolerations in Descheduler 2019-11-05 20:42:36 +00:00
Mike Dame
4b9e732c18 Switch from glog to klog 2019-10-28 20:44:02 -04:00
Kubernetes Prow Robot
dd54f1a656 Merge pull request #188 from damemi/k8s-1.16 
Bump to k8s 1.16
 2019-10-22 09:07:22 -07:00
Mike Dame
11044ed89d Run generated scripts 2019-10-21 13:33:39 -04:00
Mike Dame
c9e3c63b85 Remove codecgen 
Since [1] codecgen has been obsolete.

[1] https://github.com/kubernetes/kubernetes/issues/36120
 2019-10-21 13:33:38 -04:00
Mike Dame
b176dd2e77 Remove old NodeOutOfDisk unit test 2019-10-21 13:33:38 -04:00
Mike Dame
9ea6aa536e Update CriticalPod unit test 
The kubelet used to only determine critical pods by the namespace and an annotation (see [1], where our code switched from only checking the annotation to using the kubelet package helper, which at that revision also only checked for an annotation and namespace). However, now the kubelet uses Pod Priority to check for critical pods[3], which is controlled by an admission plugin[4]. Given that testing the functioning of the admission plugin is out of the scope of this repository, it is probably sufficient to manually set the pod priority to critical, to ensure that our Descheduler code behaves properly in response.

[1] 062b8698c7 (diff-094ac8507e224603239ce19592b518f5)
[2] https://github.com/kubernetes/kubernetes/blob/release-1.9/pkg/kubelet/types/pod_update.go#L151
[3] https://github.com/kubernetes/kubernetes/blob/a847874655/pkg/kubelet/types/pod_update.go#L153
[4] https://github.com/kubernetes/kubernetes/blob/a847874655/plugin/pkg/admission/priority/admission.go
 2019-10-21 13:33:38 -04:00
sph
5e25e21ca2 Fix golint errors 
Signed-off-by: sph <shepenghui1@huawei.com>
 2019-10-17 20:50:13 +08:00
Mike Dame
0af97c1b5e fix compile errors 2019-10-12 11:12:04 -04:00
Mike Dame
5af668e89a Remove old API install patterns 2019-10-12 11:08:46 -04:00
Mike Dame
26adf87323 Bump reactions 
Change imports to reflect current component-base usage
 2019-10-12 11:07:21 -04:00
ksimon1
fb1b5fc690 added unit test 
updated e2e test

Signed-off-by: ksimon1 <ksimon@redhat.com>
 2019-10-04 11:08:33 +02:00
ksimon1
dc7f9efc19 Descheduler can evict all types of pods with special annotation 
When pod has this annotation, descheduler will always try to evict this pod.
User can control which pods (only pods with this annotation) will be evicted.
Signed-off-by: ksimon1 <ksimon@redhat.com>
 2019-10-04 11:07:40 +02:00
Jan Chaloupka
e35eb4a0b5 Run ./hack/update-gofmt 2019-09-19 14:29:00 +02:00
Jan Chaloupka
2e6f14103b Project migrared under kubernetes-sigs, change import path prefix to sigs.k8s.io/descheduler 2019-09-19 14:09:05 +02:00