k3s/test/e2e/node
Kubernetes Submit Queue 68da45623f
Merge pull request #59217 from filbranden/cos1
Automatic merge from submit-queue. 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>.

Increase RSS limit for runtime from 300MB to 350MB on test creating 100 pods per node.

**What this PR does / why we need it**:

In recent COS image (cos-dev-65), the typical RSS increased a bit. It seems it was already close to the limit and now surpassed it.

I looked at some samples of failures of the test... In [some cases](https://k8s-gubernator.appspot.com/build/kubernetes-jenkins/logs/ci-kubernetes-e2e-gce-cosdev-k8sbeta-serial/1270) it seems very close (<1%), a [more representative example](https://k8s-gubernator.appspot.com/build/kubernetes-jenkins/logs/ci-kubernetes-e2e-gce-cosdev-k8sbeta-serial/1266)  goes up about ~5% but in [some](https://k8s-gubernator.appspot.com/build/kubernetes-jenkins/logs/ci-kubernetes-e2e-gce-cosdev-k8sbeta-serial/1269) [cases](https://k8s-gubernator.appspot.com/build/kubernetes-jenkins/logs/ci-kubernetes-e2e-gce-cosdev-k8sbeta-serial/1267) it did go up to close or above 340MB.

I did some tests using just "docker" on the machines, creating 100 "pause" containers. I didn't really see much difference in RSS, though after killing the containers RSS was a bit higher on the host running cos-dev.

I looked at some possible reasons that would explain that:
- Transparent huge pages: confirmed they were disabled;
- Version of docker: same on cos-65, cos-64 and cos-63;
- Version of Go: same, 1.9.2 on cos-64 and cos-65, while 1.9.1 on cos-63 but 64 was already using same as current version;
- Version of systemd (maybe due to systemd libraries linked into Docker?): same on the three images.

I don't really know what can explain this increase... So cc @adityakali in case he has something to add here. Also cc @yguo0905 with whom I discussed the testgrid.




**Which issue(s) this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close the issue(s) when PR gets merged)*:
N/A

**Special notes for your reviewer**:
N/A

**Release note**:

```release-note
NONE
```
2018-02-06 00:43:13 -08:00
..
BUILD Add e2e test for mount propagation 2017-11-14 10:16:30 +01:00
OWNERS
apparmor.go enable to specific unconfined AppArmor profile 2017-09-28 10:06:36 +08:00
events.go migration pod relevant e2e tests to sig node 2017-11-07 11:16:59 +08:00
framework.go Rename e2e sig framework files 2017-08-02 09:33:19 +08:00
kubelet.go fix todo: migrate to use framework.AddOrUpdateLabelOnNode/framework.RemoveLabelOffNode replace of updateNodeLabels 2018-02-03 15:52:37 +08:00
kubelet_perf.go Increase RSS limit for runtime from 300MB to 350MB on test creating 100 pods per node. 2018-02-01 12:06:54 -08:00
mount_propagation.go Review fixes 2017-11-14 10:16:30 +01:00
pod_gc.go migration pod relevant e2e tests to sig node 2017-11-07 11:16:59 +08:00
pods.go migration pod relevant e2e tests to sig node 2017-11-07 11:16:59 +08:00
pre_stop.go migration pod relevant e2e tests to sig node 2017-11-07 11:16:59 +08:00
security_context.go use core client with explicit version globally 2017-10-27 15:48:32 +08:00
ssh.go migration pod relevant e2e tests to sig node 2017-11-07 11:16:59 +08:00