Hello, since we have updated to the new slurm version (19.05) every time
a jobstep is launched with mpirun it ends with the following error message:
An ORTE daemon has unexpectedly failed after launch and before
communicating back to mpirun. This could be caused by a number
of factor
On Tuesday, 4 June 2019 6:57:39 AM PDT Jean-mathieu CHANTREIN wrote:
> Is there a way to preempt jobs by using the priority of a job calculate with
> priority_multifactor and not with a priority related to partition or qos ?
You would need to write your own preemption plugin to do that, the exist
On Monday, 3 June 2019 7:53:39 AM PDT Alexander Åhman wrote:
> That was my first thought too, but... no. Both /etc/hosts (not used) and
> slurm.conf are identical on all nodes, both working and non-working nodes.
I think Slurm caches things like that, so it might be worth restarting
slurmctld to
On Wednesday, 5 June 2019 10:04:11 AM PDT Andrés Marín Díaz wrote:
> Can it be a bug in the new version?
If it's working with srun but not with mpirun it sounds like there's some
incompatibility between how mpirun is calling srun to launch orted and what
Slurm is doing now.
You'd need to find