We are pleased to announce the availability of Slurm version 22.05.8.
This includes a number of minor to moderate bug fixes, including a
regression when updating unsorted NodeName and NodeAddr lists, and an
issue when upgrading from 21.08 with job_env or job_script set in
AccountingStoreFlags.
Slurm can be downloaded from https://www.schedmd.com/downloads.php.
- Tim
--
Tim McMullan
Release Management, Support, and Development
SchedMD LLC - Commercial Slurm Development and Support
* Changes in Slurm 22.05.8
==========================
-- Fix potential deadlock at slurmctld startup when job has invalid qos.
-- Avoid unnecessary call to clusteracct_storage_g_cluster_tres() when pinging
dynamic nodes. This avoids significant slowdowns for slurmctld when
ping_nodes() calls all nodes to re-register.
-- openapi/dbv0.0.3[6-8] - fix segfault that could arrise from Slurm database
connnection failing.
-- Fix regression introduced in 22.05.0rc1 when updating a NodeName=<nodelist>
with NodeAddr and/or NodeHostname if the specified nodelist wasn't sorted.
-- openapi/v0.0.38 - change type of nice field for job submissions to integer
from string.
-- openapi/v0.0.38 - add oversubscribe option to job submission properties.
-- openapi/v0.0.38 - fix incorrect data used to populate "time/start" field
for jobs.
-- openapi/dbv0.0.38 - avoid dumping failure if preempt qos with level 0 is
provided by slurmdbd query.
-- Avoid an fd leak when lib dir for sbcast fails to be created.
-- common/slurmdbd_pack - fix env and script hash when unpacking a 21.08
dbd_job_start_msg_t.
-- Fix a race between job_container/tmpfs, cncu, x11 setup and adoption of pids
in cray with pam_slurm_adopt.
