-
Spike
-
Resolution: Done
-
Normal
-
None
-
None
-
Upstream
-
5
-
False
-
None
-
False
-
OCPSTRAT-330 - [Upstream] OpenShift AutoScaler TechDebt (Phase 3)
-
CLOUD Sprint 239
User Story
As a developer i would like to reduce the amount of extra patches we carry in the cluster autoscaler. To help continue this effort, adjusting the provider MachineSet controllers to utilize the new annotations as proposed in the upstream autoscaler, and a deprecation plan for the old annotations, would allow me to reduce the code we carry.
Background
As defined in https://github.com/kubernetes/autoscaler/pull/4840 the cluster autoscaler will soon have scale from zero support for cluster-api. To accomplish this, some of the annotations used for denoting the scaling size of the machines in the machineset had to change. OpenShift will need to adjust its machineset controllers for each provider, and documentation, to ensure that we can use the new annotations.
Additionally, we will need to investigate and create a plan for deprecating the old annotations. Depending on the process, this might require that we carry patches allowing for both annotations.
This will require changes to the platform providers and a patch to the cluster autoscaler that we will carry until the deprecation process is complete.
Steps
- update providers MachineSet controllers to use new annotations
- create a patch for cluster autoscaler in OpenShift that will allow for both annotations
- update product documentation to display the new annotations
Stakeholders
- cloud team
Definition of Done
- cluster autoscaler continues to function as expected with respect to scaling from zero once we have rebased to use the upstream scale from zero implementation
- Docs
- product docs might need updating to display these new annotations
- Testing
- tests might need updating, this will depend on annotation usage in the tests