<div dir="ltr"><div dir="ltr">Is there a direct upgrade path from
<span style="font-size:12pt;font-family:"Times New Roman",serif">20.11.0 to 22.05.6 or is it in multiple steps?</span><br clear="all"><div><div dir="ltr" class="gmail_signature"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div><br></div><div><br></div><div>Sid Young</div><div><br></div></div></div></div></div></div></div></div></div></div><br></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Fri, Nov 11, 2022 at 7:53 AM Marshall Garey <<a href="mailto:marshall@schedmd.com">marshall@schedmd.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">We are pleased to announce the availability of Slurm version 22.05.6.<br>
<br>
This includes a fix to core selection for steps which could result in <br>
random task launch failures, alongside a number of other moderate <br>
severity issues.<br>
<br>
- Marshall<br>
<br>
--<br>
Marshall Garey<br>
Release Management, Support, and Development<br>
SchedMD LLC - Commercial Slurm Development and Support<br>
<br>
> * Changes in Slurm 22.05.6<br>
> ==========================<br>
> -- Fix a partition's DisableRootJobs=no from preventing root jobs from working.<br>
> -- Fix the number of allocated cpus for an auto-adjustment case in which the<br>
> job requests --ntasks-per-node and --mem (per-node) but the limit is<br>
> MaxMemPerCPU.<br>
> -- Fix POWER_DOWN_FORCE request leaving node in completing state.<br>
> -- Do not count magnetic reservation queue records towards backfill limits.<br>
> -- Clarify error message when --send-libs=yes or BcastParameters=send_libs<br>
> fails to identify shared library files, and avoid creating an empty<br>
> "<filename>_libs" directory on the target filesystem.<br>
> -- Fix missing CoreSpec on dynamic nodes upon slurmctld restart.<br>
> -- Fix node state reporting when using specialized cores.<br>
> -- Fix number of CPUs allocated if --cpus-per-gpu used.<br>
> -- Add flag ignore_prefer_validation to not validate --prefer on a job.<br>
> -- Fix salloc/sbatch SLURM_TASKS_PER_NODE output environment variable when the<br>
> number of tasks is not requested.<br>
> -- Permit using wildcard magic cookies with X11 forwarding.<br>
> -- cgroup/v2 - Add check for swap when running OOM check after task<br>
> termination.<br>
> -- Fix deadlock caused by race condition when disabling power save with a<br>
> reconfigure.<br>
> -- Fix memory leak in the dbd when container is sent to the database.<br>
> -- openapi/dbv0.0.38 - correct dbv0.0.38_tres_info.<br>
> -- Fix node SuspendTime, SuspendTimeout, ResumeTimeout being updated after<br>
> altering partition node lists with scontrol.<br>
> -- jobcomp/elasticsearch - fix data_t memory leak after serialization.<br>
> -- Fix issue where '*' wasn't accepted in gpu/cpu bind.<br>
> -- Fix SLURM_GPUS_ON_NODE for shared GPU gres (MPS, shards).<br>
> -- Add SLURM_SHARDS_ON_NODE environment variable for shards.<br>
> -- Fix srun error with overcommit.<br>
> -- Fix bug in core selection for the default cyclic distribution of tasks<br>
> across sockets, that resulted in random task launch failures.<br>
> -- Fix core selection for steps requesting multiple tasks per core when<br>
> allocation contains more cores than required for step.<br>
> -- gpu/nvml - Fix MIG minor number generation when GPU minor number<br>
> (/dev/nvidia[minor_number]) and index (as seen in nvidia-smi) do not match.<br>
> -- Fix accrue time underflow errors after slurmctld reconfig or restart.<br>
> -- Surpress errant errors from prolog_complete about being unable to locate<br>
> "node:(null)".<br>
> -- Fix issue where shards were selected from multiple gpus and failed to<br>
> allocate.<br>
> -- Fix step cpu count calculation when using --ntasks-per-gpu=.<br>
> -- Fix overflow problems when validating array index parameters in slurmctld<br>
> and prevent a potential condition causing slurmctld to crash.<br>
> -- Remove dependency on json-c in slurmctld when running with power saving.<br>
> Only the new "SLURM_RESUME_FILE" support relies on this, and it will be<br>
> disabled if json-c support is unavailable instead.<br>
<br>
</blockquote></div></div>