Upgrade k8s.io/* to v0.22, sigs.k8s.io/controller-runtime to v0.10

[timebertt]

How to categorize this issue?

/area open-source
/kind task

What would you like to be added:

We should upgrade to the latest versions of our go upstream dependencies:

• sigs.k8s.io/controller-runtime@v0.10.x: https://github.com/kubernetes-sigs/controller-runtime/releases/tag/v0.10.0
• k8s.io/*@v0.22.x

It's probably a good idea to complete this one before continuing on #4251.

Important changes / Action items

Here is a list of a few upstream changes to look out for when vendoring g/g, k/* and c-r in any of our repos (e.g. extensions). Please consider the release notes of controller-runtime@v0.10.0 and upwards as well for a more complete list and more details.
This list contains some action items for g/g as well.

⚠️ breaking / most notable changes

• ⚠️ Client: Ensure no stale data remains in target object kubernetes-sigs/controller-runtime#1640

  • previously, a client call wouldn't unset present fields on the in-memory object if they were unset in the API server's response (standard behavior of json decoder in stdlib / k8s libs)
  • now, the client zeroes out fields before decoding into the passed object
  • similarly, the fake client now also zeroes out fields before decoding (🐛 Fix object overwrite on patch for fake client kubernetes-sigs/controller-runtime#1651), so this might make tests using the fake client fail if they were implemented incorrectly
  • remove our workarounds in g/g for the old behavior (mainly kutil.CreateResetObjectFunc), as they are not needed anymore with the changed client behavior

• ⚠ bump to kubernetes 1.22.0 kubernetes-sigs/controller-runtime#1626

  • Many important APIs are not served anymore in 1.22, however they are still present in the published API packages (https://github.com/kubernetes/api/tree/release-1.22 and friends)
  • envtest.WebhookInstallOptions.*Webhooks and envtest.CRDInstallOptions.CRDs have to be switched from client.Object to the respective typed objects

• ⚠️ Disable the rest_client_request_latency_seconds metric by default kubernetes-sigs/controller-runtime#1587

  • currently, we don't seem to be using rest_client_request_latency_seconds (so far we are also exposing it only on a few components)

ℹ️ other changes / good to know / for information only:

• ✨Add an option to recover panic for controller reconcile kubernetes-sigs/controller-runtime#1627
  • consider setting controller.Options.RecoverPanic=true -> Set RecoverPanic to true for controller reconcile #5383
• ✨ add health check for webhook server kubernetes-sigs/controller-runtime#1588
  • consider using server.StartedChecker() for components serving webhooks -> Add ready check for Webhook server #5380
• 🐛 envtest: get gvk from hook struct instead of forcing set TypeMeta kubernetes-sigs/controller-runtime#1665
  • don't need to set GVK / TypeMeta on envtest.WebhookInstallOptions.*Webhooks explicitly anymore
• 🌱 Source should retry to get informers until timeout expires  kubernetes-sigs/controller-runtime#1678
  • only applies to native controller-runtime controllers using source.Kind, g/g can only benefit from it once Refactor Gardener components to native controller-runtime components #4251 is completed
• 🌱 Disable fsync on etcd envtest kubernetes-sigs/controller-runtime#1629
  • upgrade envtest binaries to at least v1.21 (to get etcd v3.5) and make sure, we use --unsafe-no-fsync to stop writing to disk in envtests ➡️ Bump envtest binaries to v1.22 #4851
• 🌱 Manager should use the global logger, reduce log names verbosity kubernetes-sigs/controller-runtime#1647

[timebertt]

I will wait for #4757 progress before starting on this in order to avoid too many conflicts.

[timebertt]

/assign

[timebertt]

/unassign
as I have completed all steps that were on my list. The other remaining tasks in this list are not urgent and can be taken over by others as well.

[rfranzke]

/assign @acumino

[acumino]

@timebertt, IIUC consider the setting controller.Options.RecoverPanic=true can only be done for the controller, native to controller-runtime. If that is right, I think we should move ahead with this after the refactoring of gardener-controller-manager and gardenlet. WDYT?

[timebertt]

can only be done for the controller, native to controller-runtime

Yes, that's true.

We can either do it for the controller's that are already controller-runtime native controllers today (admission-controller, resource-manager, scheduler) or later once all controllers are refactored.
However I expect "later" to be rather in a few months because of other priorities. So I would vote for doing it now for our c-r native controllers.

[acumino]

/cc @rfranzke
Can we close this given all the extensions PR are merged now?

[rfranzke]

Sure, please close if there is nothing left to be done :)

[acumino]

Sure, please close if there is nothing left to be done :)

Ready check PR for all provider-extension I think there is no need for this in any other extension.

[acumino]

/close

[gardener-prow]

@acumino: Closing this issue.

In response to this:

/close

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.