-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
Hi folks,
Slurm, like Torque, sets RLIMIT_DATA to match the mem request of a job.
However, on Linux malloc() in glibc will use mmap() for any
non-trivial allocation and mmap() obeys RLIMIT_AS and not RLIMIT_DATA.
So Slurm's setting has no effect
Could you tell how is the program foo built? Which MPI version and PMI
library are you using?
在 2013-07-17三的 13:16 -0700,Ralph Castain写道:
Hi folks
We're trying to test the pmi2 support in 2.6.0 and hitting a problem.
We have verified that the pmi2 support was built/installed, and that
We are currently planning on removing accounting_storage/pgsql and
accounting_storage/filetxt in the next Slurm release. Both are out of
date and we feel not widely used.
If anyone really wants either of these plugins to stay around speak now
or forever hold your peace.
If anyone is using
My $.02,
Filetxt is simplest to use/setup, portable ( non DB dependent ), and has a
higher chance of not failing ie, database not being available, certain
flavors of DB going in and out of style like postgres (as exampled by your
choice to no longer support it ), I say keep it.
We currently use
Hello Danny,
We've been using filetxt for its simplicity. We have some perl scripts
that are used to generate some reports. Since its not causing a
problem, I'd vote for keeping it. If not, I can still add it myself,
and compile the new versions..
~ Seren
On Thu, Jul 18, 2013 at 7:47 PM,
It was built using Open MPI, which was configured with
set SLURMLOC = /opt/slurm; ./configure --with-slurm=$SLURMLOC
--with-pmi=$SLURMLOC --with-verbs --without-mpi-param-check
--enable-orterun-prefix-by-default --prefix=$PWD/install
--with-mxm=/opt/mellanox/mxm --with-fca=/opt/mellanox/fca
On 2013-07-18 19:18, Danny Auble wrote:
We are currently planning on removing accounting_storage/pgsql and
accounting_storage/filetxt in the next Slurm release. Both are out of
date and we feel not widely used.
We are still using accounting_storage/filetxt, because it was easiest to
setup.
You might be able to alter the moab_2_slurmdb.pl found in
contribs/slurmdb-direct to read in your acct file and put it in the
dbd. If you do please submit it and we will include it in the same
directory for others.
Danny
On 07/18/13 10:40, Rolandas Naujikas wrote:
On 2013-07-18 19:18,
You might be able to alter the moab_2_slurmdb.pl found in
contribs/slurmdb-direct to read in your acct file and put it in the
dbd. If you do please submit it and we will include it in the same
directory for others.
Danny
On 07/18/13 10:40, Rolandas Naujikas wrote:
On 2013-07-18 19:18,
I am trying to use a Login node, other then the primary slurm scheduler
node, for batch job submission but it does not work.
on the master node:
fe1 ~]$ sbatch sbatch
Submitted batch job 59
on any batch node:
gpu-ln1 ~]$ sbatch sbatch
sbatch: error: Batch job submission failed:
Basically the main reason we want to get rid of the filetxt plugin is it
is becoming hard to maintain and some people in testing have found
issues that have always existed but don't in the mysql plugin. Since it
doesn't normally make since to use this filetxt plugin in production,
and
Yes, I just tried that. Despite AllocNodes is for executing jobs
(I just want to submit them to the queue/partition to be executed
later) it seems AllocNodes is needed for submit to a queue as well.
Thanks
Eva
On Thu, 18 Jul 2013, Seren Soner wrote:
Perhaps you could lookup AllocNodes in
12 matches
Mail list logo