Add support for NetworkUnavailable node condition

[nicolas-goudry]

What would you like to be added:

Set the NetworkUnavailable condition on Nodes when the CNI comes up or goes down.

Why is this needed:

With this condition, it would be easier to tell when there is an issue with the CNI.

---

Reference discussion

[KR411-prog]

I want to work on this issue. Could you assign it to me?

[jdn5126]

@KR411-prog thank you for the help! It is fine to work on while unassigned, feel free to post a PR whenever or lmk if you want to discuss

[KR411-prog]

sure..

[github-actions]

This issue is stale because it has been open 60 days with no activity. Remove stale label or comment or this will be closed in 14 days

[github-actions]

This issue is stale because it has been open 60 days with no activity. Remove stale label or comment or this will be closed in 14 days

[github-actions]

This issue is stale because it has been open 60 days with no activity. Remove stale label or comment or this will be closed in 14 days

[dougbyrne]

Please unstale

[github-actions]

This issue is stale because it has been open 60 days with no activity. Remove stale label or comment or this will be closed in 14 days

[bawejahritik]

I'd like to work on this, can you please point me to the starting point

[antoineDievDecath]

Hey @jdn5126,

Can you tell us if this feature is in your roadmap ?
Since latest versions VPC-CNI is taking more time to start, sometimes I can see pods be scheduled and started before aws-node became totally ready.
In my opinion pod shouldn't be scheduled on node as long as the VPC-CNI isn't ready.

[jdn5126]

@antoineDievDecath this issue is not currently on our roadmap, and I do not believe it will be anytime soon, unfortunately.

For pod scheduling, the node will not be marked as ready until the CNI conflist is copied to /etc/cni/net.d/10-aws.conflist. When upgrading the aws-node image, there will be a period where pods are waiting for the CNI to finish initialization, since the conflist already exists.

I am surprised to hear you say the latest version is taking more time to startup. Which versions are you comparing? Internally, we see the startup time has greatly decreased in the latest versions. It should be less than a second.

[jchen6585]

This issue cannot be fixed through aws-node pod as VPC-CNI does not have permissions to patch the node for security reasons. https://github.com/aws/amazon-vpc-cni-k8s/blob/master/config/master/aws-k8s-cni.yaml#L317

[github-actions]

⚠️COMMENT VISIBILITY WARNING⚠️

Comments on closed issues are hard for our team to see.
If you need more assistance, please open a new issue that references this one.
If you wish to keep having a conversation with other community members under this issue feel free to do so.