v1.92.0

What's Changed

Key New Features 🎉

• feat: add ML Diagnostics module and integration for GKE TPU blueprints by @AdarshK15 in #5350
• NAP support on GKE Clusters (gke-cluster module) by @SwarnaBharathiMantena in #5420
• feat: optional infra setup for inference gateway by @jessicaochen in #5453
• feat(slurm): support compact placement with DWS Flex-Start for H4D, A3Ultra and A4 by @parulbajaj01 in #5579

Breaking Changes 🚨

• Transitioning to Slurm Native Auth with resilient workbench keys distribution by @arpit974 in #5695
• default to sauth for newer deployments in h4d and a3mega-gcsfuse blueprints by @arpit974 in #5707

New Modules 🧱

• adding new dns-managed-zone module. by @arpit974 in #5485
• adding new global static ip module. by @arpit974 in #5559
• adding new module for kubernetes namespace. by @arpit974 in #5562
• adding new iap-policy module. by @arpit974 in #5564
• Adding new cloud run module. by @arpit974 in #5567
• adding new redis module. by @arpit974 in #5569
• adding new kubernetes-secret module. by @arpit974 in #5572
• adding new workload_identity_binding module. by @arpit974 in #5574
• adding new scripting module gke-backend-fetcher under community folder. by @arpit974 in #5593
• adding a new helm-upgrade module under community folder. by @arpit974 in #5595
• adding new spanner-migrations runner module under community folder. by @arpit974 in #5597

Module Improvements 🔨

• Adding native K8s annotations and GKE cluster enhancements by @arpit974 in #5610
• Default Kueue config for Pathways by @scaliby in #5628

Improvements 🛠

• [Telemetry] Get blueprint even from deployment directory by @kadupoornima in #5656
• [Telemetry] Capture exit code upon fatal command failures by @kadupoornima in #5658
• (gke) Remove additional network settings from A3U blueprint by @agrawalkhushi18 in #5652
• (gke) Remove additional networks from A4 and A4X family blueprints by @agrawalkhushi18 in #5682
• (gke) Remove additional network settings from TPU v6e,7x and g4 by @agrawalkhushi18 in #5692
• [Telemetry] Add support to merge vars from deployment files and CLI --vars by @kadupoornima in #5694
• [Telemetry] Add support for collection of CPU machines and Default machines when unset in module by @kadupoornima in #5696
• Make Managed lustre default in A3u and A3m series Slurm blueprints by @saara-tyagi27 in #5396
• [Telemetry] Add a retry mechanism to get the GCP Project information to eliminate transient issues by @kadupoornima in #5702
• [Telemetry] Add an atomic flag to ensure telemetry event is not recurrently called by @kadupoornima in #5705
• Pin DCGM to version 4.5.3 by @shubpal07 in #5721
• feat(gke): expose monitoring components as a parameter by @cboneti in #5722
• feat(job submission): Dynamic topology routing for gke jobs by @Neelabh94 in #5664

Deprecations 💤

• Remove hpc-slurm-static blueprint by @kadupoornima in #5672

Version Updates ⏫

• Fix A3 HighGPU test by pinning GKE version to 1.33 to resolve COS incompatibility by @kadupoornima in #5673
• Update minimum required Packer version to 1.15.3 by @AdarshK15 in #5701

Bug fixes 🐞

• fix: Add tpu_topology conditional logic for TPU flex start by @agrawalkhushi18 in #5655
• fix: Update the vpc module output name for additional network by @agrawalkhushi18 in #5690
• fix(slurm): correct vNUMA socket and SMT thread calculations in util.py by @kadupoornima in #5683
• Multi NIC support & cluster ID fix for Slurm controller by @rahimkhan19 in #5563
• [Telemetry] Collect the correct exit code when user intentionally stops deployment (0 instead of 1) by @kadupoornima in #5704
• Clean up custom spot VM variables during standard fallback by @rahimkhan19 in #5697
• fix: accelerator label auto resolution by @Neelabh94 in #5717

Full Changelog: v1.91.0...v1.92.0