You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The current implementation of the hashing mechanism that is utilized for drift does not make any consideration when there are updates to the NodePool CRDs during karpenter upgrades. The problem that is present here is that all nodes on a customers cluster will get drifted without any changes to the node configuration, but when upgrading the Karpenter CRDs.
More generally, Karpenter will need guard against the following cases during a karpenter upgrade:
NodePool adds a default value to existing fields that are included in hash calculations
NodePool field added to the hash calculation with an already-set values
NodePool field, with and without defaults values, is removed from the hash calculations
Without safeguards, users that upgrade to new versions of Karpenter with changes to Drift hashing mechanism will see their whole fleet rolled over and recycled.
Description
What problem are you trying to solve?
The current implementation of the hashing mechanism that is utilized for drift does not make any consideration when there are updates to the NodePool CRDs during karpenter upgrades. The problem that is present here is that all nodes on a customers cluster will get drifted without any changes to the node configuration, but when upgrading the Karpenter CRDs.
More generally, Karpenter will need guard against the following cases during a karpenter upgrade:
Without safeguards, users that upgrade to new versions of Karpenter with changes to Drift hashing mechanism will see their whole fleet rolled over and recycled.
This would required to support in solving the following issues: #909, #337, #493, and aws/karpenter-provider-aws#5447
How important is this feature to you?
This will need to be implemented before v1
The text was updated successfully, but these errors were encountered: