Improve shutdown logic: Wait until no requests are made

[motoki317]

What this PR does / why we need it:

Hello!
This PR mainly introduces 3 changes:

• Improves shutdown logic of the controller.
  • The controller now waits until no more requests are seen to the nginx process before exiting. This is done by polling /nginx_status (stub_status) page, and see if the requests handled number goes up.
  • resolves Zero downtime upgrade #6928.
• Changes the grace-shutdown-period in chart values to a more reasonable default.
  • Kubernetes pods are terminated (sent a SIGTERM) and removed from Service endpoints in parallel, so pods are expected to wait a bit before stopping to accept new connections from the upstream - this is the very reason this PR exists. It is generally a good idea in Kubernetes to wait a little bit after receiving a SIGTERM signal before exiting - I have changed the default wait time from 0s to 10s.
  • resolves Zero downtime upgrade #6928 along with the above change.
• Removes the unnecessary (verbose) wait-shutdown binary and preStop hook.
  • This preStop hook seemed to serve no purpose - see Distinguish wait-shutdown command from standard k8s SIGTERM #6287 for discussion.
  • resolves Distinguish wait-shutdown command from standard k8s SIGTERM #6287.

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• CVE Report (Scanner found CVE and adding report)
• Breaking change (fix or feature that would cause existing functionality to change)
  • I should note that: this PR technically introduces a breaking change to the default configuration values, but shouldn't be much of a problem to the users.
• Documentation only

Which issue/s this PR fixes

fixes #6928
fixes #6287

How Has This Been Tested?

• I have built docker image on this branch and manually tested the improved shutdown logic feature.
  • See https://gist.github.com/motoki317/902637f6761e289098925d7db1999f43 for how I tested it
• I have added one E2E test spec to verify that the improved shutdown logic indeed works. I have run all E2E tests locally and verified that all tests pass.

Checklist:

• My change requires a change to the documentation.
• I have updated the documentation accordingly.
• I've read the CONTRIBUTION guide
• I have added unit and/or e2e tests to cover my changes.
• All new and existing tests passed.

[linux-foundation-easycla]

The committers listed above are authorized under a signed CLA.

• ✅ login: motoki317 / name: motoki317 (872d64b, e76ea47, 25f5d09, ff45a2a, ae42b6a, 3fe18f1, 4f43d4a)

[k8s-ci-robot]

Welcome @motoki317!

It looks like this is your first PR to kubernetes/ingress-nginx 🎉. 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/ingress-nginx 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]

Hi @motoki317. Thanks for your PR.

I'm waiting for a kubernetes member to verify that this patch is reasonable to test. If it is, they should reply with /ok-to-test on its own line. Until that is done, I will not automatically test new commits in this PR, but the usual testing commands by org members will still work. Regular contributors should join the org to skip this step.

Once the patch is verified, the new status will be reflected by the ok-to-test label.

I understand the commands that are listed here.

Details

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-sigs/prow repository.

[netlify]

✅ Deploy Preview for kubernetes-ingress-nginx canceled.

Name	Link
🔨 Latest commit	ae42b6a
🔍 Latest deploy log	https://app.netlify.com/sites/kubernetes-ingress-nginx/deploys/6747dc3c140560000844fa9c

[adrianmoisey]

/ok-to-test

[Gacko]

Please do not change the manifests in deploy/static/. This is part of the release process.

[motoki317]

@Gacko I see, I've reverted the changes there.

[motoki317]

Hi, I'm wondering what's the status of this PR.
Is this being discussed or reviewed,
or should I be a bit more patient and wait?

[longwuyuan]

Reviewed. But too much going on so please wait. Thank you.

…

On Fri, 29 Nov, 2024, 06:44 motoki317, ***@***.***> wrote: Hi, I'm wondering what's the status of this PR. Is this being discussed or reviewed, or should I be a bit more patient and wait? — Reply to this email directly, view it on GitHub <#12397 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABGZVWRZK3OK5BIUMXQRSHL2C65YHAVCNFSM6AAAAABSG6HKRWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDKMBWHEYTCMZYG4> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[strongjz]

/kind feature
/priority backlog
/triage accepted

I have added this to the 1.13 release milestone.

[strongjz]

lgtm, ill bring this up with @Gacko

/hold

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: motoki317, strongjz

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

The pull request process is described here

Details

Needs approval from an approver in each of these files:

• OWNERS [strongjz]

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

[dcw329]

We really need this. please. if 1.13 is not coming because of ingate, can we get this put into a patch for 1.12. we have google lb's sending traffic to dead controller pods in production causing 521 on cloudflare, and HTTP probe failed with statuscode: 500 on kubernetes. We are getting dropped traffic.

[rikatz]

/uncc

[motoki317]

@dcw329 I believe you can set longer grace-shutdown-period to cope with this problem in the latest release. This PR only adds a "better workaround" and better default values.
But let's bring futher discussion to kubernetes/kubernetes#106476 (or somewhere else in this repo) to keep this thread clean.

@Gacko Hello, any updates with the hold status? (I know the team is busy, but I don't think a friendly reminder would hurt.) :)

[rittneje]

@motoki317 I see two things that are concerning in this PR.

1. It hard-codes the assumption that 5 seconds is enough time to conclude that no more requests are coming from the load balancer, and does not offer any command line flag to configure this.
2. Now shutdown-grace-period is not applied until after nginx stops receiving requests (assuming it ever does). And that in turn means that NGINX's graceful shutdown for existing connections is now arbitrarily delayed as long as any connection continues to receive requests. There needs to be a way to skip over the new polling loop.

In addition, I'm not sure how this polling loop improves anything. The problem we've seen in AWS is that it can take up to three minutes for the pod to stop receiving new connections from the NLB once it starts draining. In other words, polling for "no new requests" in a 5 second window does not mean anything if the NLB can still route new connections to the pod. That is why we need shutdown-grace-period to delay the actual nginx shutdown via SIGQUIT.

[k8s-ci-robot]

PR needs rebase.

Details

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-sigs/prow repository.