Before this PR we would make demands for the entire application: The driver pod + all executor pods. This was a problem because the driver pod also existed in the cluster as well as the demand which was needed to provide scale up signal in a particular AZ.
After this PR
This PR adds the driver pod to PodNamesByNamespace which allows scaling infrastructure to to deduplicate driver pods and their associated demands
==COMMIT_MSG==
Deduplicate driver pod demands
==COMMIT_MSG==
Before this PR
Before this PR we would make demands for the entire application: The driver pod + all executor pods. This was a problem because the driver pod also existed in the cluster as well as the demand which was needed to provide scale up signal in a particular AZ.
After this PR
This PR adds the driver pod to
PodNamesByNamespace
which allows scaling infrastructure to to deduplicate driver pods and their associated demands==COMMIT_MSG== Deduplicate driver pod demands ==COMMIT_MSG==