Exception for sig-autoscaling Configurable Tolerance feature

[Jean-Marc François]

• Enhancement name: Configurable Tolerance
• Enhancement status (alpha/beta/stable): alpha
• SIG: autoscaling
• k/enhancements repo issue #: 4951
• PR #’s: 130797
• Additional time needed (in days): 2
• Reason this enhancement is critical for this milestone: This long-overdue enhancement for sig-autoscaling will allow the community to tune the threshold under which pod autoscaling ignores fluctuations of the metric used for autoscaling. This threshold is currently hard-coded to 10%, which is (much) too high for large deployments (where 10% maps to a large number of pods), and for users that need fast autoscaling reactions. 
• Risks from adding code late: (to k8s stability, testing, etc.) Since the updated logic is behind a feature gate, this should be little risk.
• Risks from cutting enhancement: (partial implementation, critical customer usecase, etc.) Users handling large deployments or users that need fast reaction time will continue facing poor pod autoscaling performances.

[Jean-Marc François]

All,

The PR corresponding to this exception has been merged: https://github.com/kubernetes/kubernetes/pull/130797

Since it was merged before the exception deadline, it can be included in v1.33.

(Please keep me informed if there's still something missing.)

Thank you for granting this exception, much appreciated!

Jean-Marc

jm-franc@