Table of contents

What's new and changed in the scheduling service

The scheduling service release and subsequent refreshes can include new features, bug fixes, and security updates. Refreshes appear in reverse chronological order, and only the refreshes that contain updates for the scheduling service are shown.

You can see a list of the new features for the platform and all of the services at What's new in IBM Cloud Pak for Data?

Installing or upgrading the scheduling service

Ready to install or upgrade the scheduling service?

Related documentation:

Refresh 6 of Cloud Pak for Data Version 3.5

A new version of the scheduling service was released in May 2021.

Assembly version: 1.1.4

This refresh includes the following changes:

Bug fixes

Issue: The scheduling service cannot schedule pods on nodes that are tainted by the Db2® service. Therefore, Db2 pods that needed to be placed on the tainted nodes remain in pending state.

Resolution: The scheduling service can now schedule pods on any nodes that have been tainted.

Refresh 5 of Cloud Pak for Data Version 3.5

A new version of the scheduling service was released in April 2021.

Assembly version: 1.1.3

This refresh includes the following changes:

Bug fixes

Issue: Agent pods cannot start on some ppc64le clusters due to insufficient memory.

Resolution: The pod memory was increased.

Refresh 3 of Cloud Pak for Data Version 3.5

A new version of the scheduling service was released in February 2021.

Assembly version: 1.1.2

This release includes the following changes:

Bug fixes
Issue: The network policy prevented the scheduling service from communicating with the OpenShift® API server on IBM Cloud.

Resolution: The scheduling service can now communicate with the OpenShift API server.

Issue: Prometheus metrics were incorrectly rounded.

Resolution: The metrics are rounded correctly.

Issue: Prometheus metrics for historical resource use data were reported incorrectly.

Resolution: The metrics are reported correctly.

Issue: Too few GPUs were reclaimed to support consumers with insufficient resources.

Resolution: The correct number of GPUs are reclaimed and allocated to underfed consumers.

Issue: When a node is cordoned, pods are incorrectly preempted.

Resolution: This issue is resolved.

Issue: Pods and parallel jobs remained in the pending state with the message:
Kubernetes pods corresponding to this job do not exist.

Resolution: The scheduling service detects that the pods have been created and attempts to schedule them.

Refresh 2 of Cloud Pak for Data Version 3.5

A new version of the scheduling service was released in January 2021.

Assembly version: 1.1.1

This release includes the following changes:

New features

You must install Version 1.1.1 of the scheduling service if you want to install the service on Red Hat® OpenShift 4.6.

In addition this release also includes the following features and updates:

Quota enforcement
If you want to programmatically enforce the quotas that you set for Cloud Pak for Data or for various Cloud Pak for Data services, you must install Version 1.1.1 of the scheduling service on your cluster.

For details on quota enforcement, see Managing the platform.

For details on installing the scheduling service, see Setting up the scheduling service.

Co-scheduling of pods
(For Watson™ Machine Learning Accelerator.) The ability to co-schedule pods is used by parallel and AI workloads in Watson Machine Learning Accelerator to:
  • Guarantee that all pods can start
  • Remove resource deadlock
  • Enable workloads to grow and shrink
  • Support reclaiming pods in the event of resource contention
Improved GPU sharing
(For Watson Machine Learning Accelerator.) The scheduling service allows GPUs to be shared between competing groups, which improves GPU utilization. Sharing policies govern how to resolve resource contention.

Initial release of Cloud Pak for Data Version 3.5

The scheduling service was released as part of Cloud Pak for Data Version 3.5.

Assembly version: 1.1.0

The scheduling service is required if you plan to install the Watson Machine Learning Accelerator service.