We are pleased to announce the availability of Slurm version 24.05.7.
This release fixes some stability issues in 24.05, including a crash in
slurmctld after updating a reservation with an empty nodelist.
Downloads are available at https://www.schedmd.com/downloads.php .
--
Marshall Garey
Release Management, Support, and Development
SchedMD LLC - Commercial Slurm Development and Support
> * Changes in Slurm 24.05.7
> ==========================
> -- Fix slurmctld crash when after updating a reservation with an empty
> nodelist. The crash could occur after restarting slurmctld, or if
> downing/draining a node in the reservation with the REPLACE or REPLACE_DOWN
> flag.
> -- Fix jobs being scheduled on higher weighted powered down
> nodes.
> -- Fix memory leak when RestrictedCoresPerGPU is enabled.
> -- Prevent slurmctld deadlock in the assoc mgr.
We are pleased to announce the availability of Slurm version 24.11.3.
24.11.3 fixes the database cluster ID generation not being random, a
regression in which slurmd -G gave no output, a long-standing crash in
slurmctld after updating a reservation with an empty nodelist, and some
other minor to moderate bugs.
Downloads are available at https://www.schedmd.com/downloads.php .
--
Marshall Garey
Release Management, Support, and Development
SchedMD LLC - Commercial Slurm Development and Support
> * Changes in Slurm 24.11.3
> ==========================
> -- Fix race condition in slurmrestd that resulted in "Requested
> data_parser plugin does not support OpenAPI plugin" error being returned
> for valid endpoints.
> -- If multiple partitions are requested, set the SLURM_JOB_PARTITION
> output environment variable to the partition in which the job is running
> for salloc and srun in order to match the documentation and the behavior of
> sbatch.
> -- Fix regression where slurmd -G gives no output.
> -- Don't print misleading errors for stepmgr enabled steps.
> -- slurmrestd - Avoid connection to slurmdbd for the following
> endpoints:
> GET /slurm/v0.0.41/jobs
> GET /slurm/v0.0.41/job/{job_id}
> -- slurmrestd - Avoid connection to slurmdbd for the following
> endpoints:
> GET /slurm/v0.0.40/jobs
> GET /slurm/v0.0.40/job/{job_id}
> -- Significantly increase entropy of clusterid portion of the
> sluid by seeding the random number generator
> -- Avoid changing process name to "watch" from original daemon name.
> This could potentially breaking some monitoring scripts.
> -- Avoid slurmctld being killed by SIGALRM due to race condition
> at startup.
> -- Fix slurmctld crash when after updating a reservation with an empty
> nodelist. The crash could occur after restarting slurmctld, or if
> downing/draining a node in the reservation with the REPLACE or REPLACE_DOWN
> flag.
> -- Fix race between task/cgroup cpuset and jobacctgather/cgroup.
> The first was removing the pid from task_X cgroup directory causing
> memory limits to not be applied.
> -- srun - Fixed wrongly constructed SLURM_CPU_BIND env variable
> that could get propagated to downward srun calls in certain mpi
> environments, causing launch failures.
> -- slurmrestd - Fix possible memory leak when parsing arrays with
> data_parser/v0.0.40.
> -- slurmrestd - Fix possible memory leak when parsing arrays with
> data_parser/v0.0.41.
> -- slurmrestd - Fix possible memory leak when parsing arrays with
> data_parser/v0.0.42.
