Dynamically set Disk Volume size.

[gurpalw]

Description

What problem are you trying to solve?

Karpenter provisions nodes of all sizes, which is great. But this means a 50GB root volume which may be sufficient for most nodes, will not be sufficient for the larger nodes as more pods can run therefore more images are downloaded.

Karpenter should be able to dynamically set the root volume size based on the size of node, using rules set by the karpenter user.

How important is this feature to you?
This will prevent us from excessive spend, as the alternative is to simply increase the nodeclasses volume size. this is inefficient since we share a EC2NodeClass across many different node type/sizes. It will also prevent DiskPressure evictions.

• Please vote on this issue by adding a 👍 reaction to the original issue to help the community and maintainers prioritize this request
• Please do not leave "+1" or "me too" comments, they generate extra noise for issue followers and do not help prioritize the request
• If you are interested in working on this issue or have submitted a pull request, please leave a comment

[jonathan-innis]

+1, this idea looks quite similar to aws/karpenter-provider-aws#2394 in the AWS provider repo -- one thought here is if we just capture the total pod requests for the ephemeral storage, then we could use that to scale the volume size -- the only issue that you might run into there is that you might end up having CPU/memory leftover on the instance and that will go unused -- we could potentially think about assigning some "default" ephemeral storage for every Gi or CPU that is going unused to create a little headroom on the instance

[jonathan-innis]

/triage accepted

[jonathan-innis]

/priority backlog