Tags in managed node groups

[derrickburns]

Tags (managedNodeGroups[*].tags) in managed nodegroups apply to the EKS Nodegroup resource and do not propagate to the provisioned Autoscaling Group like in unmanaged nodegroups.

Does this mean that tags will not get applied to instances and that the cluster autoscaler will not work?

[whereisaaron]

There is supposedly built-in support for cluster-autoscaler tags. This will only work if AWS automatically tags the ASG for the node group taints and labels, without then you can't scale from zero.

https://docs.aws.amazon.com/eks/latest/userguide/managed-node-groups.html

Amazon EKS tags managed node group resources so that they are configured to use the Kubernetes Cluster Autoscaler.

Important
If you are running a stateful application across multiple Availability Zones that is backed by Amazon EBS volumes and using the Kubernetes Cluster Autoscaler, you should configure multiple node groups, each scoped to a single Availability Zone. In addition, you should enable the --balance-similar-node-groups feature.

[johnpetersjr]

I have experienced that when setting tags on a NodeGroup, the tags will propagate to the ASG and to the EC2 instances.

When settings tags on a managedNodeGroup, the tags do NOT propagate down to the ASG or EC2.

[sashasimkin]

Is there a way to tag corresponding ASGs and instances?

[galindro]

According to https://docs.aws.amazon.com/eks/latest/userguide/eks-using-tags.html, Amazon EKS managed node groups do not supports tag propagation yet. Currently, we would need to do some code to find the associated ASGs and tag them

[TBBle]

aws/containers-roadmap#608 is the AWS containers roadmap issue tracking this feature.

[TBBle]

For Cluster Autoscaler specifically, the EC2 instances and Auto Scaling Groups created for Managed NodeGroups are automatically tagged with the expected labels:

k8s.io/cluster-autoscaler/enabled:true
k8s.io/cluster-autoscaler/<clustername>: owned

[github-actions]

This issue is stale because it has been open 30 days with no activity. Remove stale label or comment or this will be closed in 5 days.

[github-actions]

This issue is stale because it has been open 30 days with no activity. Remove stale label or comment or this will be closed in 5 days.

[TBBle]

It seems the recommended way to apply tags to EC2 instances in EKS Managed Node Groups is using Launch Templates, which eksctl supports. Is that sufficient to fulfill this feature request?

Cluster Autoscaler doesn't depend on tags on the EC2 instances, it depends on tags on the ASG, which are still unable to be customised per aws/containers-roadmap#608.

[michaelbeaumont]

Seems like we could just set the tags on the launch template, since eksctl always uses launch templates

[aclevername]

The managedNodeGroups[*].tags field already gets added to the launch template spec:
https://github.com/weaveworks/eksctl/blob/da899ad2176936d1045dc365ea3bba44647858c1/pkg/cfn/builder/managed_launch_template.go#L145-L164

This adds the tag to all EC2 instances. You still cannot add tags to the ASG aws/containers-roadmap#608 (comment)

[aclevername]

I'm going to close this issue as the original intent of this was to be able to tag the EC2 instances, which you can.

It seems the recommended way to apply tags to EC2 instances in EKS Managed Node Groups is using Launch Templates, which eksctl supports. Is that sufficient to fulfill this feature request?

Cluster Autoscaler doesn't depend on tags on the EC2 instances, it depends on tags on the ASG, which are still unable to be customised per aws/containers-roadmap#608.

I think if/when EKS add support for tagging the autoscaler it will likely be a no-op from our end, as we are already tagging the nodegroup and launch template, so its likely that if it was ever implemented the ASG would inherit the tags from one of those two.

[afirth]

@aclevername Without this change, cluster-autoscaler scale up from 0 (eksctl docs) doesn't work with managed node groups.
Is it that eksctl can't implement this, or doesn't prioritize? If the latter, are you open to community contributions?
Either way, I'd like to reopen this for either a docs update to make it clear that it only works with unmanaged node groups, or a fix. The upstream issue is at 500 votes and still in "researching" - unlikely to be implemented I guess

[aclevername]

are you open to community contributions?

absolutely!

Either way, I'd like to reopen this for either a docs update to make it clear that it only works with unmanaged node groups, or a fix. The upstream issue is at 500 votes and still in "researching" - unlikely to be implemented I guess

👍

[aclevername]

@Himangini

[Himangini]

@aclevername isn't this related to what @Skarlso is working on here #4853 🤔 maybe I am missing something.

@afirth thanks for your comments, we are happy to accept PRs if you have a fix in mind 🙌🏻 🎉

[cPu1]

Tags (managedNodeGroups[*].tags) in managed nodegroups apply to the EKS Nodegroup resource and do not propagate to the provisioned Autoscaling Group like in unmanaged nodegroups.

Please note that this behaviour has changed to:

Tags (managedNodeGroups[*].tags) in managed nodegroups apply to the EKS Nodegroup resource and to the EC2 instances launched as part of the nodegroup.
They do not propagate to the provisioned Autoscaling Group like in unmanaged nodegroups.

[afirth]

#5002 is open for review to fix this.
@cPu1 the problem with that is that cluster autoscaler is looking at the ASG tags

[afirth]

I think #5002 didn't link in a way to autoclose, but this is implemented there and should be closed?

[Skarlso]

Yep. Thanks! :)