adding in priority value in podSpec for controller, scheduler and apiserver

[erleene]

We have found a bug in 1.13.x where mirror pods get evicted when the node is under pressure. This PR sets the priority directly so that those pods won't get evicted. We don't believe that this is required with 1.14+.

Relates to this issue kubernetes/kubernetes#73572

[k8s-ci-robot]

Welcome @erleene!

It looks like this is your first PR to kubernetes-incubator/kube-aws 🎉. Please refer to our pull request process documentation to help your PR have a smooth ride to approval.

You will be prompted by a bot to use commands during the review process. Do not be afraid to follow the prompts! It is okay to experiment. Here is the bot commands documentation.

You can also check if kubernetes-incubator/kube-aws has its own contribution guidelines.

You may want to refer to our testing guide if you run into trouble with your tests not passing.

If you are having difficulty getting your pull request seen, please follow the recommended escalation practices. Also, for tips and tricks in the contribution process you may want to read the Kubernetes contributor cheat sheet. We want to make sure your contribution gets all the attention it needs!

Thank you, and welcome to Kubernetes. 😃

[k8s-ci-robot]

Thanks for your pull request. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

📝 Please follow instructions at https://git.k8s.io/community/CLA.md#the-contributor-license-agreement to sign the CLA.

It may take a couple minutes for the CLA signature to be fully registered; after that, please reply here with a new comment and we'll verify. Thanks.

---

• If you've already signed a CLA, it's possible we don't have your GitHub username or you're using a different email address. Check your existing CLA data and verify that your email is set on your git commits.
• If you signed the CLA as a corporation, please sign in with your organization's credentials at https://identity.linuxfoundation.org/projects/cncf to be authorized.
• If you have done the above and are still having issues with the CLA being reported as unsigned, please log a ticket with the Linux Foundation Helpdesk: https://support.linuxfoundation.org/
• Should you encounter any issues with the Linux Foundation Helpdesk, send a message to the backup e-mail support address at: login-issues@jira.linuxfoundation.org

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. I understand the commands that are listed here.

[codecov-io]

Codecov Report

Merging #1631 into v0.13.x will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff            @@
##           v0.13.x    #1631   +/-   ##
========================================
  Coverage    25.49%   25.49%           
========================================
  Files           98       98           
  Lines         5049     5049           
========================================
  Hits          1287     1287           
  Misses        3619     3619           
  Partials       143      143

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 289cd68...9d45c31. Read the comment docs.

[davidmccormick]

Please test before we merge

[dominicgunn]

This looks to have been fixed for Kubernetes 1.13.5+, the default for kube-aws is now 1.13.7 so this shouldn't impact newly created clusters.

Kubernetes upstream has fixed this for versions as far back as v1.11.x, do we want to do the same in kube-aws?

[davidmccormick]

Yes, that has got us confused somewhat as we observed the behavior on a 1.13.7 cluster without this work-around! It shouldn’t happen.

[erleene]

@dominicgunn

The problem still occurs in 1.13.7, where no controllers are running and getting evicted. We have the api server and the scheduler up, and kubelet shows the following:


11:01:46.328766    1525 certificate_manager.go:378] Certificate request was not signed: timed out waiting for the condition
11:01:46.328804    1525 certificate_manager.go:269] Reached backoff limit, still unable to rotate certs: timed out waiting for the condition
11:10:49.829455    1525 log.go:172] http: multiple response.WriteHeader calls
11:14:54.907091    1525 certificate_manager.go:378] Certificate request was not signed: timed out waiting for the condition




some more interesting output:

10:49:05.306713    1525 reflector.go:125] k8s.io/kubernetes/pkg/kubelet/kubelet.go:444: Failed to list *v1.Service: Get https://127.0.0.1/api/v1/services?li>
10:49:05.307641    1525 reflector.go:125] k8s.io/kubernetes/pkg/kubelet/config/apiserver.go:47: Failed to list *v1.Pod: Get https://127.0.0.1/api/v1/pods?fi>
10:49:05.308780    1525 reflector.go:125] k8s.io/kubernetes/pkg/kubelet/kubelet.go:453: Failed to list *v1.Node: Get https://127.0.0.1/api/v1/nodes?fieldSel>

[dominicgunn]

Does this only happen in clusters created with kube-aws, or is it a problem in upstream k8s?

[davidmccormick]

Thanks, it was also confirmed that the controller-manager pods had been evicted - this is an upstream kubernetes issue which may also affect other releases, e.g. 1.14.x, 1.15.x (although it is supposed to have been patched). It affects kube-aws customers deploying smaller controllers that are pushed near or over capacity and won't affect well sized nodes. What it has shown us though is that when resources are under pressure some of the most critical components were chosen for eviction.

I suggest we merge this as a work-around on 1.13.x and test the behaviour on 1.14.x and apply similar work-around if observed on that branch.

Good work Erleen!

[dominicgunn]

Cool, i'm behind that. Code looks good.

[dominicgunn]

/approve

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: dominicgunn

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [dominicgunn]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[davidmccormick]

/lgtm

[davidmccormick]

Thanks for your first contribution! 🙏