v1.53.5

(internal-only release)

v1.53.4

(internal-only release)

v1.53.3

Added

• Support for job-routing plugins, from @dposada

v1.53.2

Added

• Features to compute clusters, from @dposada

Changed

• Constrained checkpointing to supported pools, from @nsinkov

v1.53.1

Fixed

• Fix the names for synthetic pod workload labels, from @dposada

v1.53.0

Added

• Environment variables for telemetry, from @dposada

Changed

• Improved performance for VirtualMachineLeaseAdapter, TaskRequestAdapter, and update-host-reservation, from @scrosby

Fixed

• Hard delete pods that have been in the terminating state for too long, from @dposada
• Skip inactive pools when ranking, from @dposada

v1.52.0

Added

• Instance field with how long the job queued before that instance, from @dposada

Changed

• Made straggler kill a mea-culpa failure, from @nsinkov

v1.51.14

Changed

• Added pod labels for application name and version, from @dposada
• Prefixed all application pod labels with the configured pod label prefix, from @dposada

v1.51.13

Fixed

• Calculate the time-until-waiting metric correctly, from @scrosby

v1.51.12

Changed

• Do fenzo unassigns in batches outside of the k8s state locks, from @scrosby
• Make the k8s lock vector a vector not a sequence, from @scrosby
• Split metrics for synthetic pods and regular pods in k8s, from @scrosby