Custom Drift Interval

[Robyt96]

Description

What problem are you trying to solve?
Hi, I would like some clarification on the drift mechanism of Karpenter. From this post and from some tests conducted, I observe that Karpenter has a drift interval of approximately 5 minutes. Here an exctract from the logs:

{"level":"INFO","time":"2024-02-29T10:20:20.902Z","logger":"controller.disruption","message":"disrupting via drift replace, terminating 1 candidates ip-IP.REGION.compute.internal/r5.xlarge/on-demand and replacing with on-demand node from types r5.xlarge","commit":"1072d3b"}
{"level":"INFO","time":"2024-02-29T10:25:43.811Z","logger":"controller.disruption","message":"disrupting via drift replace, terminating 1 candidates ip-IP.REGION.compute.internal/r5.xlarge/on-demand and replacing with on-demand node from types r5.xlarge","commit":"1072d3b"}
{"level":"INFO","time":"2024-02-29T10:30:12.311Z","logger":"controller.disruption","message":"disrupting via drift replace, terminating 1 candidates ip-IP.REGION.compute.internal/r5.xlarge/on-demand and replacing with on-demand node from types r5.xlarge","commit":"1072d3b"}
{"level":"INFO","time":"2024-02-29T10:36:25.142Z","logger":"controller.disruption","message":"disrupting via drift replace, terminating 1 candidates ip-IP.REGION.compute.internal/r5.xlarge/on-demand and replacing with on-demand node from types r5.xlarge","commit":"1072d3b"}
{"level":"INFO","time":"2024-02-29T10:40:13.588Z","logger":"controller.disruption","message":"disrupting via drift replace, terminating 1 candidates ip-IP.REGION.compute.internal/r5.xlarge/on-demand and replacing with on-demand node from types r5.xlarge","commit":"1072d3b"}
{"level":"INFO","time":"2024-02-29T10:44:12.187Z","logger":"controller.disruption","message":"disrupting via drift replace, terminating 1 candidates ip-IP.REGION.compute.internal/r5.xlarge/on-demand and replacing with on-demand node from types r5.xlarge","commit":"1072d3b"}
{"level":"INFO","time":"2024-02-29T10:47:40.626Z","logger":"controller.disruption","message":"disrupting via drift replace, terminating 1 candidates ip-IP.REGION.compute.internal/r5.xlarge/on-demand and replacing with on-demand node from types r5.xlarge","commit":"1072d3b"}
{"level":"INFO","time":"2024-02-29T10:51:49.059Z","logger":"controller.disruption","message":"disrupting via drift replace, terminating 1 candidates ip-IP.REGION.compute.internal/r5.xlarge/on-demand and replacing with on-demand node from types r5.xlarge","commit":"1072d3b"}
{"level":"INFO","time":"2024-02-29T10:56:31.900Z","logger":"controller.disruption","message":"disrupting via drift replace, terminating 1 candidates ip-IP.REGION.compute.internal/r5.xlarge/on-demand and replacing with on-demand node from types r5.xlarge","commit":"1072d3b"}

Is it currently possible to specify this drift interval with any parameters, or is this something that could be implemented?

How important is this feature to you?
Our organization requires us to regularly change the AMIs of the cluster nodes, and the drift feature of Karpenter is very useful in this regard. However, our nodes host Java/Springboot workflows with pods that are sometimes very slow to start, so we would need to wait longer between one drift replacement and another.

• 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

[engedaam]

Since kubernetes-sigs/karpenter isn't aware of provider-specific NodeClass, it doesn't track watch events for these NodeClass. We will soon be able to just watch v1beta1.NodeClaimReference objects to make drift occur immediately. Closing this issue as a duplicate of #337. If you feel that this issue would not solve your use case, please feel free to re-open!

[engedaam]

/close

[k8s-ci-robot]

@engedaam: 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.