Cluster Toolkit v1.108.0 adds Slurm support for TPU v5e, v5p, v6e and gcluster explain

Cluster Toolkit v1.108.0 is available. This release lets you pin node pools to reservation sub-blocks in the gke-node-pool module, adds Slurm support for Cloud TPU v5e, v5p, and v6e, and introduces the gcluster explain command to inspect blueprints and modules.

This release also exposes the wait_for_pods_ready configuration for Kueue in the kubectl-apply module and sets the recovery_timeout field to 0s by default for Pathways workloads.

Additionally, this release suppresses configuration drift for clusters and node pools in the ignore_changes argument of the lifecycle block for Google Kubernetes Engine (GKE), and upgrades the network modules for the Cloud Foundation Toolkit (CFT) to version 18.3 and later.

This update also checks credentials for Artifact Registry and Docker only when a job submission uses Crane to build container images.

For more information, see the release announcement on GitHub.