Overview
SUNK v8.2.0 was released on October 23, 2026. It ships with Slurm25.11.8-coreweave.2 by default, up from 25.11.8-coreweave.1 in SUNK 8.1. See SUNK to Slurm version mapping.
The main additions in 8.2 are gang scheduling for Kubernetes Jobs and LeaderWorkerSets through the SUNK Pod Scheduler, and a move from a generated topology.conf to a generated topology.yaml. The release also limits out-of-memory (OOM) kills in login and compute Pods to a single process, adds zsh to the login image, and changes several defaults.
Before you upgrade, review Changes to check before you upgrade.
Upgrade to 8.2
- SUNK Self-Service
- SUNK Standard
Set
sunkVersion on the SunkCluster resource. To also take the Slurm build that ships with 8.2, set slurmVersion in the same change. The following command upgrades a cluster in the tenant-slurm Namespace, the Namespace the Cloud Console uses, to SUNK 8.2.0 and its default Slurm build. Replace CLUSTER_NAME with the name of your SunkCluster resource.slurmVersion requires the leading v. sunkVersion works with or without it. For how the rollout treats running jobs, see Upgrade a SUNK Self-Service cluster.Changes to check before you upgrade
New in 8.2
- Gang scheduling for Jobs and LeaderWorkerSets: The SUNK Pod Scheduler schedules the Pods of a Kubernetes Job, or of each LeaderWorkerSet replica group, together as one Slurm job. As a result, a workload starts only when all of its Pods can run. No annotation is needed. The scheduler automatically cancels orphaned placeholder jobs, which
scheduler.config.scheduler.orphanJobSweepcontrols. See Run workloads with gang scheduling. - Login Pod topology spread: Login Pods now spread across nodes by default, with
maxSkew: 1onkubernetes.io/hostname. Setlogin.topologySpreadConstraintsto replace the default, or set it to[]to turn it off. The change takes effect when a login Pod is recreated. - zsh on login Pods: The login image now includes zsh. SUNK’s aliases,
ulimitandumasksettings, and thedrainandundrainhelpers load in both bash and zsh. To make zsh a user’s shell, setnsscache.defaultShell: /bin/zsh, or set the user’ssunkLoginShellattribute (SCIM) orloginShellattribute (LDAP). Slurm command completion remains bash-only. See Set the default shell. - OOM kill containment: When a login or
slurmdcontainer runs out of memory, the kernel now kills a single process instead of the whole container, and SSH sessions are killed first. On compute nodes, jobs are killed beforeslurmd, so the container doesn’t restart. Compute containment requires cgroup v2. - NodeSet conditions and rendered spec: NodeSets report a
PodSpecValidcondition when the API server rejects a compute Pod, and aComputeConfigValidcondition when theSlurmClustercompute settings are invalid. The operator no longer patches NodeSetspec. It writes the effective Pod template to the read-onlysunk.coreweave.com/renderedannotation instead, so GitOps tools stop reporting drift on NodeSets. Manual edits to operator-derived fields inspec, such as theslurmdimage, are now ignored instead of reverted. - Clearer disk drain reasons: When the health check drains a node because
/or/tmpis full, the drain reason starts withNHC: check_cw_fs_used:and lists the space that/tmp,/coredumps, and/opt/sunkuse, largest first. The thresholds are unchanged. - Inode lock fixer on by default: New
SlurmClusterresources enable the inode lock fixer for the managed database (database.mysqlCluster.inodeLockFixerin theslurmchart). Existing clusters keep their current setting.
Fixes
- A running job submitted with more than one QoS keeps the QoS it started with when
slurmctldrestarts or reconfigures. Before 8.2, it could switch to the highest-priority QoS. - SUNK now sets
RealMemoryfrom theslurmdcontainer’s memory limit, minus reserved memory, instead of the node type’s hardware memory. Full-node jobs on NodeSets whose memory limit is lower than the node type’s memory now schedule. - SUNK can now set a node stuck in
COMPLETINGtoDOWN. Before 8.2, its cleanup controller failed withslurm_update error: Invalid user id. - The
reconfigure-watch,sync-users-from-system, andnsscache-updatesidecars exit cleanly on shutdown. Before 8.2, they were killed at the end of the grace period, which delayedslurmctldteardown. - When
compute.ssh.enabledisfalse, the operator deletes the computesshds6 ConfigMaps. - Argo CD no longer reports a permanent diff on the
sunkcontroller manager Deployment.