NodeCondition on failing to run system pods

[SergeyKanzhelev]

There are many cases when Node become broken when system pods are failing to run. For example:

• Image pull issues
• Init container failures
• Other Crash Loop Backoff cases

I think introducing a new Node Condition when any of the system namespace pods has failed will be very useful.

Seeking for feedback - are there ever cases when system pods are failing to run and node is not "degraded".

Also, may be a rule like this may be configurable to include additional namespaces to check.

[SergeyKanzhelev]

/kind feature

[chenggu88]

Generally I think it will be useful to expose that node is degraded without all the system component running .

Are the system components defined as running in kube-system namespace or priority=system-cluster-critical or system-node-critical?

[SergeyKanzhelev]

Are the system components defined as running in kube-system namespace or priority=system-cluster-critical or system-node-critical?

I thought about the namespace. Priority is also a good idea. It is just namespace is often more controlled so less chance to have an "unexpected" DaemonSet in that namespace, while priority may be easier to set. But we can do both

[SergeyKanzhelev]

Thinking about how to imlpement this, I thhink we will need kubernetes/enhancements#4188 for the best implementation

[k8s-triage-robot]

The Kubernetes project currently lacks enough contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue as fresh with /remove-lifecycle stale
• Close this issue with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle stale