From 822a27f9afde052b7b1e452f96be0b3d0fa6adf6 Mon Sep 17 00:00:00 2001 From: Moe Jette <jette1@llnl.gov> Date: Fri, 8 Aug 2008 22:34:30 +0000 Subject: [PATCH] Add information about memlock limit propagation --- doc/html/faq.shtml | 29 +++++++++++++++++++++++++++-- 1 file changed, 27 insertions(+), 2 deletions(-) diff --git a/doc/html/faq.shtml b/doc/html/faq.shtml index 33784e0e5cb..9061ba99828 100644 --- a/doc/html/faq.shtml +++ b/doc/html/faq.shtml @@ -23,8 +23,10 @@ name for a batch job?</a></li> allocated to a SLURM job?</a></li> <li><a href="#terminal">Can tasks be launched with a remote terminal?</a></li> <li><a href="#force">What does "srun: Force Terminated job" indicate?</a></li> -<li><a href="#early_exit">What does this mean: "srun: First task exited 30s ago" -followed by "srun Job Failed"?</a></li> +<li><a href="#early_exit">What does this mean: "srun: First task exited +30s ago" followed by "srun Job Failed"?</a></li> +<li><a href="#memlock">Why is my MPI job failing due to the locked memory +(memlock) limit being too low?</a></li> </ol> <h2>For Administrators</h2> <ol> @@ -485,9 +487,32 @@ not normally productive. This behavior can be changed using srun's period or disable the timeout altogether. See srun's man page for details. +<p><a name="memlock"><b>18. Why is my MPI job failing due to the +locked memory (memlock) limit being too low?</b></a><br> +By default, SLURM propagates all of your resource limits at the +time of job submission to the spawned tasks. +This can be disabled by specifically excluding the propagation of +specific limits in the <i>slurm.conf</i> file. For example +<i>PropagateResourceLimitsExcept=MEMLOCK</i> might be used to +prevent the propagation of a user's locked memory limit from a +login node to a dedicated node used for his parallel job. +If the user's resource limit is not propagated, the limit in +effect for the <i>slurmd</i> daemon will be used for the spawned job. +A simple way to control this is to insure that user <i>root</i> has a +sufficiently large resource limit and insuring that <i>slurmd</i> takes +full advantage of this limit. For example, you can set user's root's +locked memory limit limit to be unlimited on the compute nodes (see +<i>"man limits.conf"</i>) and insuring that <i>slurmd</i> takes +full advantage of this limit (e.g. by adding something like +<i>"ulimit -l unlimited"</i> to the <i>/etc/init.d/slurm</i> +script used to initiate <i>slurmd</i>). +Related information about <a href="#pam">PAM</a> is also available. + <p class="footer"><a href="#top">top</a></p> + <h2>For Administrators</h2> + <p><a name="suspend"><b>1. How is job suspend/resume useful?</b></a><br> Job suspend/resume is most useful to get particularly large jobs initiated in a timely fashion with minimal overhead. Say you want to get a full-system -- GitLab