Description
Queue resource calculations are not centralised, scattered through many classes and a source of dozens of bugs. We need a common place to store these calculations, simplify them if possible and extend it with proper logging.
Example of such calculations:
AbstractLeafQueue#assignContainers
AbstractCSQueue#accept
AbstractLeafQueue#calculateAndGetAMResourceLimitPerPartition
AbstractLeafQueue#updateMaximumApplications
... and many more.
Attachments
Issue Links
- Dependent
-
YARN-10965 Centralize queue resource calculation based on CapacityVectors
- Resolved
- relates to
-
YARN-10888 [Umbrella] New capacity modes for CS
- Resolved