That was it.

Thanks for the help.
Jeff

From: Morris Jette [mailto:[email protected]]
Sent: Thursday, March 05, 2015 7:19 PM
To: slurm-dev
Subject: [slurm-dev] Re: Odd problem with CPU totals

Perhaps a memory limit?
On March 5, 2015 2:57:41 PM PST, "Sarlo, Jeffrey S" 
<[email protected]<mailto:[email protected]>> wrote:

I am not sure what I am missing, but we are having problems submitting jobs 
that request all the cpus on a node.  This is an example

$ srun -N 1 -n 4 -t 2:00:00 --pty /bin/bash -l
srun: error: Unable to allocate resources: Requested node configuration is not 
available
$ srun -N 1 -n 3 -t 2:00:00 --pty /bin/bash -l
[compute-0-0]$

Each of the nodes in the cluster has 2 sockets and 2 cores per socket for a 
total of 4 cpus.
# ssh compute-0-0 slurmd -C
ClusterName=(null) NodeName=compute-0-0 CPUs=4 Boards=1 SocketsPerBoard=2 
CoresPerSocket=2 ThreadsPerCore=1 RealMemory=3957 TmpDisk=8570 UpTime=0-00:43:42

I also have this in my slurm.conf

#
# COMPUTE NODES
#
#NodeName=DEFAULT  CPUs=4 SocketsPerBoard=2 CoresPerSocket=2 ThreadsPerCore=1 
RealMemory=3957 State=UNKNOWN Weight=1
NodeName=DEFAULT CPUs=4 RealMemory=3957 State=UNKNOWN Weight=1

I have tried both lines

  and

get the same result.

What would keep the first job request from being able to run?

Thanks.
Jeff

--
Sent from my Android device with K-9 Mail. Please excuse my brevity.

Reply via email to