v1.71.0

What's Changed

Module Improvements 🔨

• Adding validations for naming resources by @vikramvs-gg in #4788

Improvements 🛠

• Add Managed Lustre support in gke-a4x blueprint by @parulbajaj01 in #4793

Bug fixes 🐞

• fix munge mount on login failure due to slow ctrl slurmv6 setup by @m4r1k in #4702

New Contributors

• @m4r1k made their first contribution in #4702

Full Changelog: v1.70.0...v1.71.0

v1.70.0

What's Changed

Breaking Changes 🚨

• Removing support for maintenance_interval for reservations created by TAMs by @LAVEEN in #4748
• Migration of jobset from static manifests to helm chart and upgrading version to 0.10.1 by @shubpal07 in #4765

Module Improvements 🔨

• Add automated TPU support and GCS integration in TPU v6 blueprint by @shubpal07 in #4755

Improvements 🛠

• H4d blueprint refactored by @rachit-google in #4740

Full Changelog: v1.69.0...v1.70.0

v1.69.0

What's Changed

Key New Features 🎉

• Add NUMA-aware scheduling in GKE clusters (enabled for G4) by @kadupoornima in #4760
• Add daily PR integration tests for G4 machines by @kadupoornima in #4761

New Modules 🧱

• Add module providing mount scripts for WEKA filesystems by @wiktorn in #3509

Improvements 🛠

• Adding GKE sample for running nvidia-bug-report by @raushan2016 in #4741
• PSA update by @okrause in #4744

New Contributors

• @aslam-quad made their first contribution in #4742

Full Changelog: v1.68.0...v1.69.0

v1.68.0

What's Changed

Key New Features 🎉

• downloading libnccl2 and libnccl-dev for a3u and a4h by @rachit-google in #4680

Breaking Changes 🚨

• Allowing setting use_job_duration with non-exclusive partitions. by @arpit974 in #4696
• Add multi-network support in TPU v6e by @agrawalkhushi18 in #4723
• Update vpc and cloud_router versions in VPC network module by @kadupoornima in #4732

Module Improvements 🔨

• Refactoring in gke persistent module by @vikramvs-gg in #4618
• Migrate Kueue installation to use Helm chart by @shubpal07 in #4542

Improvements 🛠

• Update nvidia DRA driver version to v25.3.0 by @parulbajaj01 in #4670
• Updated A3-mega and A4-high Slurm blueprints to adopt nvidia add repository scirpt. by @rachit-google in #4667
• Update H4D blueprint: disable automatic updates, provide image info, and delete duplicate filestore by @Neelabh94 in #4644
• Add Managed Lustre support in gke-a4 by @parulbajaj01 in #4654
• Add Managed Lustre support in gke a3 ultra by @parulbajaj01 in #4700
• Adds an irdma health check to h4d nodes by @samskillman in #4704
• Enable Spot VM Provisioning For H4D by @LAVEEN in #4735
• Add slurm-gke blueprint by @ACW101 in #4607

Version Updates ⏫

• Updating google provider version upper bound to 7.6.0 - latest version by @bytetwin in #4736

Bug fixes 🐞

• Remove superfluous addition of chs logs to cloud ops config by @abbas1902 in #4679
• Adding "datacenter-gpu-manager-4-dev" as an additional installation in A* YAML files. by @Neelabh94 in #4623
• minor bug fix on MFT version comparison by @ljqg in #4689
• Fix inconsistent plan on Slurm cluster reconfigure by @wiktorn in #4538
• Update process to filter out starting comments in a source yaml file by @SwarnaBharathiMantena in #4707
• Fix gke build failures by @annuay-google in #4708
• Update machine-leaning/a3-ultragpu-8g/nemo-framework to fix segmentation fault error by @SwarnaBharathiMantena in #4725

New Contributors

• @mufaqam-gcl made their first contribution in #4688
• @wtempel made their first contribution in #4705
• @nikosavola made their first contribution in #4720
• @ACW101 made their first contribution in #4607

Full Changelog: v1.67.0...v1.68.0

Release v1.67.0

What's Changed

Key New Features 🎉

• Adding Support for aarch64 Based Architecture by @ljqg in #4637

Module Improvements 🔨

• added nvidia-repositories script by @rachit-google in #4553

Improvements 🛠

• Install NCCL/gIB .deb and .rpm packages for A3U and A4 by @rachit-google in #4543
• updating example to use jax ai images by @pulasthi in #4575
• Enabling Spot VM For A3 Mega/High by @LAVEEN in #4634

New Contributors

• @pulasthi made their first contribution in #4575

Full Changelog: v1.66.0...v1.67.0

Release v1.66.0

What's Changed

Key New Features 🎉

• H4D enable gcsfuse and set cluster availability type to ZONAL by @kadupoornima in #4608
• Add G4 GKE base blueprints by @kadupoornima in #4560

Module Improvements 🔨

• Slinky upgraded to v0.3.1 by @sharabiani in #4548
• Update Managed lustre gke blueprint by @parulbajaj01 in #4603

Improvements 🛠

• Making separate integration test for nccl test in gke a3 ultra by @shubpal07 in #4622
• Upgrade to Slurm 25.05 by @LAVEEN in #4606
• Hotfix: H4D Blueprint provisioning model option update by @abbas1902 in #4640

New Contributors

• @saara-tyagi27 made their first contribution in #4619

Full Changelog: v1.65.0...v1.66.0

v1.65.1: Hotfix: H4D Blueprint provisioning model options update

What's Changed

Improvements 🛠

• Hotfix: H4D Blueprint provisioning model options update by @abbas1902 in #4640

Full Changelog: v1.65.0...v1.65.1

v1.65.0

What's Changed

Improvements 🛠

• Surface Managed Lustre support in a4x by @RachaelSTamakloe in #4576
• Expand A* gpu network wait solution by @RachaelSTamakloe in #4584
• Restart slurmctld.service before scontrol reconfigure by @RachaelSTamakloe in #4609
• Support use of other shared file locations for NCCL Tests by @RachaelSTamakloe in #4615
• Add sudo to systemctl restart by @RachaelSTamakloe in #4626

Deprecations 💤

• Deprecate Debian blueprints from a3 mega gpu by @rachit-google in #4537

Bug fixes 🐞

• Power down non-responding node if there is not instance attached by @abbas1902 in #4627

Full Changelog: v1.64.0...v1.65.0

Release v1.64.0

What's Changed

Key New Features 🎉

• GKE Managed Lustre integration by @vikramvs-gg in #4572

Breaking Changes 🚨

• updated the storage for a3Ultra to basic ssd by @rachit-google in #4516

Improvements 🛠

• Allow alternative services for private-service-access by @okrause in #4477

Version Updates ⏫

• Revert gke-node-pool module to using google-beta provider by @kadupoornima in #4577

New Contributors

• @okrause made their first contribution in #4477

Full Changelog: v1.63.0...v1.64.0

Release v1.63.0

What's Changed

Key New Features 🎉

• Switch to 6-11 slurm image versions by @abbas1902 in #4425

Breaking Changes 🚨

• updated file storage to basic hdd for A3ULTRA. by @rachit-google in #4511
• Update vm boot disk and coupling it with VM instance by @arpit974 in #4558

Module Improvements 🔨

• Improvements applied to gke-pv-module by @sharabiani in #4499
• Add support for regional Instance Template in Slurm scripts by @sharabiani in #4501
• Added partial Slinky install option by @sharabiani in #4508
• Improvements on gke-node-pool module by @sharabiani in #4509
• Add GPU type to Slurm's gres.config by @sharabiani in #4546
• Add GKE Nodes support to Slurm controller by @sharabiani in #4547

Improvements 🛠

• add a4x image build recipe by @RachaelSTamakloe in #4495
• Transition DWS Flex-Start to Regional MIGs by @abbas1902 in #4491

Bug fixes 🐞

• Fix: Slurm scripts incorrectly identifying all accelerators as GPUs by @sharabiani in #4498
• Get assuredCount via specificReservation by @abbas1902 in #4540
• Fix filter for gke-nodepool instance_templates by @sharabiani in #4559

New Contributors

• @Neelabh94 made their first contribution in #4552

Full Changelog: v1.62.2...v1.63.0