Nick Tran
Nick Tran
How have you come to these numbers? Did you test them out yourself? I can understand that these timeouts shouldn't be one size fits all, and it sounds like the...
Hey @thelabdude, this is a good callout of core differences in CA's deprovisioning and Karpenter's deprovisioning. Karpenter intentionally has chosen to not use a threshold, as for any threshold you...
Seems like this issue is asking for the feature add of "toggle budget default behavior", where the current behavior is that when no budget is active, it's maximally permissive (meaning...
Can you share your nodepools and your pod specs that you're expecting to be scheduled? Can you share the full logs too?
Can you share your NodePools? Can you also share your logs?
Hey @shubhamrai1993, sorry for the extensive delay on the response here. You're right in that our current algorith does: 1. look at the existing buckets (nodes) 2. look at the...
This log should be better in more recent versions. Can you try upgrading to the latest to see if that fits your requirement?
Ah gotcha. I edited the issue title to be more reflective of the concern. Let me see if I can make this better.
Thanks for the feature request! Drifted nodes should be deprovisioned serially, and we don't deprovision a node if the needed replacement node doesn't initialize and become ready. I have a...