Commit Graph

11 Commits

Author SHA1 Message Date
Da K. Ma
793ed98715 Added more UT for invalid case.
Signed-off-by: Da K. Ma <klaus1982.cn@gmail.com>
2018-04-25 11:14:24 +08:00
Kubernetes Submit Queue
98968c36d8
Merge pull request #62715 from bsalamat/fix_antiaffinity
Automatic merge from submit-queue (batch tested with PRs 62761, 62715). 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 inter-pod anti-affinity check to consider a pod a match when all the anti-affinity terms match

**What this PR does / why we need it**:
Inter-pod anti-affinity check used to incorrectly consider a pod a match when any of the anti-affinity terms matched the pod. This PR fixes the logic to consider a pod a match when all the terms match the pod.

**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 #62567

**Special notes for your reviewer**:

**Release note**:

```release-note
Fix inter-pod anti-affinity check to consider a pod a match when all the anti-affinity terms match.
```

/sig scheduling
2018-04-17 22:38:13 -07:00
Bobby (Babak) Salamat
c590ec7ae9 Fix an issue in inter-pod affinity predicate that cause affinity to self being processed incorrectly 2018-04-14 19:46:38 -07:00
Bobby (Babak) Salamat
4d7fff1257 Add test to ensure anti-affinity matches against all terms 2018-04-14 19:32:58 -07:00
Bobby (Babak) Salamat
418c7502f0 Improve performance of affinity/anti-affinity predicate 2018-04-11 21:17:29 -07:00
Rohit Agarwal
87dda3375b Delete in-tree support for NVIDIA GPUs.
This removes the alpha Accelerators feature gate which was deprecated in 1.10.
The alternative feature DevicePlugins went beta in 1.10.
2018-04-02 20:17:01 -07:00
Rohit Agarwal
e6db88b12d Resources prefixed with *kubernetes.io/ should remain unscheduled if they are not exposed on the node.
Currently, resources prefixed with *kubernetes.io/ get scheduled to any
node whether it's exposing that resource or not.

On the other hand, resources prefixed with someother.domain/ don't get
scheduled to a node until that node is exposing that resource (or if the
resource is ignored because of scheduler extender).

This commit brings the behavior of *kubernetes.io/ prefixed resources in
line with other extended resources and they will remain unscheduled
until some node exposes these resources.

This also includes renaming IsDefaultNamespaceResource() to
IsNativeResource().
2018-03-28 17:24:30 -07:00
Yang Guo
8d880506fe Support cluster-level extended resources in kubelet and kube-scheduler
Co-authored-by: Yang Guo <ygg@google.com>
Co-authored-by: Chun Chen <chenchun.feed@gmail.com>
2018-02-27 17:25:30 -08:00
tossmilestone
3fdacfead5 Fix golint errors in pkg/scheduler based on golint check 2018-02-08 15:22:47 +08:00
Yongkun Anfernee Gui
68c2c79362 Refactor HostIP predicate algorithm
- Remove string decode logic. It's not really helping to find the
  conflict ports, and it's expensive to do encoding/decoding
- Not to parse the container ports information in predicate meta, use
  straight []*v1.ContainerPort
- Use better data structure to search port conflict based on ip
  addresses
- Collect scattered source code into common place
2018-01-08 17:10:21 -08:00
Jonathan Basseri
30b89d830b Move scheduler code out of plugin directory.
This moves plugin/pkg/scheduler to pkg/scheduler and
plugin/cmd/kube-scheduler to cmd/kube-scheduler.

Bulk of the work was done with gomvpkg, except for kube-scheduler main
package.
2018-01-05 15:05:01 -08:00