Hi,
Is it possible to specify the GRES type with MaxTRESPerNode? E.g:
slurm.conf
(…)
GresType=gpu
(…)
NodeName=gpu[01,02] NodeAddr=10.2.1.[1,2] Gres=gpu:tesla:2,gpu:kepler:2 ...
(…)
PartitionName=gpu_one Nodes=gpu[01,02] QOS=part_gpu_one …
PartitionName=gpu_two Nodes=gpu[01,02] QOS=part_gpu_two
Hi,
I have a question regarding the knl_generic plugin. Of all our nodes only 4 are
KNL nodes (in their own partition), but it seems that the syscfg tool must be
installed also on the non-KNL nodes, otherwise the log files are filled with
messages like:
error: node_features_p_node_state:
Hi,
With the possibility of rebooting KNL nodes in different MCDRAM modes
(flat,cache,…) I am wondering what is the recommended approach to set
RealMemory in slurm.conf for those nodes to always be able to use all the
available memory?
Here some background information:
In flat mode,
Thank you for the input!
@Ryan
I didn’t’ check the code of the knl_generic plugin, but I would expect that in
this case it should not represent an unexpected reboot from Slurm’s point of
view.
@Costin
Increasing ResumeTimeout in slurm.conf worked. Although slurmd is still down
after reboot,
Hi,
We want to introduce Intel Knights Landing (KNL) nodes into our cluster, and
observed the following problem: Node reboots successfully with desired NUMA and
MCDRAM modes, but remains in state down.
From slurmctl.log:
(…)[
2017-05-16T15:16:10.437] _update_node_avail_features: nodes
Hi all
Background:
I’m new to Slurm. I Installed Slurm 16.05 (Munge auth, MariaDB managed by
slurmdbd). Users maintained by LDAP, no LDAP user imported into Sturm DB so far.
bash$ sacctmgr show user
User Def Acct Admin
- -- -