2025.01 (January 2025)

Deployment Notes and What's New

When updating to a new version of the software, be sure to review both the Deployment Notes and the What’s New topics. The Deployment Notes alert you to changes to the deployment process. The What’s New topics can include critical changes, such as changes to system requirements and pod names. For both the Deployment Notes and the What’s New topics, be sure to review each note after the version that is currently deployed, up to and including the version to which you are updating.

What's New topics with critical changes are identified by "Critical Change" in the title. Deployment Notes are located in SAS Viya Platform: Deployment Notes.

Critical Change: Update Procedure for SAS with SingleStore

A SAS Viya platform software update that includes SAS with SingleStore no longer requires manual steps to pause the SingleStore cluster unless the deployment and updates are performed using Kubernetes commands.

For most deployment methods, the procedure to update deployments that include SAS with SingleStore has been streamlined. For software updates using the SAS Viya Platform Deployment Operator or the sas-orchestration command, a manual step is no longer required to pause the SingleStore cluster. This pre-update step has been automated for those deployment methods.

This additional automation also enhances the sas-stop-all and sas-start-all CronJobs. Starting with 2025.01, if you use CronJobs to stop and start a SAS Viya platform deployment, the SingleStore pods are also managed by the CronJobs. When you execute the sas-stop-all CronJob, it pauses SingleStore, and when you execute the sas-start-all CronJob, it restarts SingleStore. Previously, you were required to pause SingleStore manually before executing the sas-start-all CronJob.

The documentation has been updated to reflect this enhancement.

CAS Server Enhancement

The CAS server has been enhanced with a new management feature for MPP CAS deployments that include an optional graphics processing unit (GPU). To take advantage of additional capabilities that are provided by a GPU, CAS workers require scheduling to GPU-enabled nodes in the CAS node pool.

Previously, any resource requests or limits that you applied to nodes using a transformer always applied equally to the CAS controller and to all CAS worker nodes. These settings were always applied to the CAS node pool rather than to individual machines. You can now configure distinct node pools for CAS controller pods and CAS worker pods. With enhancements to the SAS GPU reservation service, CAS controller pods can be scheduled on nodes without GPUs, while CAS worker pods are scheduled on nodes that provide the more expensive GPU resources.

The SAS GPU reservation service enables GPU resource sharing by SAS processes. For more information about the service, see the README titled "SAS GPU Reservation Service," located at $deploy/sas-bases/examples/gpu/README.md (for Markdown format) or at $deploy/sas-bases/docs/sas_gpu_reservation_service.htm (for HTML format).

For instructions about setting up distinct node pools for CAS controllers and CAS workers, including configuration steps, see the README file located at $deploy/sas-bases/overlays/cas-server/auto-resources/README.md (for Markdown format) or at $deploy/sas-bases/docs/auto_resources_for_cas_server_for_sas_viya.htm (for HTML format).