v1.62.1

Added

• Add support for pool quotas across pools, from @scrosby

v1.62.0

Added

• Add support for routing jobs between pools based on constraints, from @scrosby

v1.61.3

Updating changelog and version for 1.61.3

v1.61.2

Changed

• Add more opentracing spans to the match cycle, @samincheva

Fixed

• Disabled pools integration tests handle 0 quota better, from @samincheva

v1.61.1

Added

• Opentracing for the match cycle logic, from @samincheva

Fixed

• Allow preemptions for tasks with unknown status, from @ahaysx
• Fix rebalancer integration test to handle failures better, from @ahaysx

v1.61.0

Changed

• Moved the global launch/kill ordering lock to be per compute-cluster, from @laurameng

v1.60.2

Fixed

• Fix bug in api-only flag that would fail operations requiring a connection to the leader, from @samincheva

v1.60.1

Changed

• Cook now determines which pool a k8s node is in via a label instead of a taint, from @scrosby

Fixed

• Ascribe NodeAffinity k8s failures to node preemption, from @scrosby

v1.59.7

Changed

• Updated instance->user cache to handle fake entities for waiting jobs, to speed up worst-case rank loop performance, by @scrosby

Fixed

• Updated structured logging utility to handle failed json conversion, from @samincheva
• Updated format-map-for-structured-logging to traverse nested maps instead of flattening them, from @samincheva

v1.59.6

Changed

• Updated scheduler, compute cluster, and kubernetes API code to use structured logging, from @samincheva