This web mail archive is frozen.
This page is part of a frozen web archive of this mailing list.
You can still navigate around this archive, but know that no new mails
have been added to it since July of 2016.
Click here to be taken to the new web archives of this list; it includes all the mails that are in this frozen archive plus all new mails that have been sent to the list since it was migrated to the new archives.
I've done a good bit of reading and still cannot get my mpirun programs to
execute without giving warnings about the memory lock limit not being
I'm on an infiniband-(openib)-networked cluster under slurm resource mgmt
Initially I did not want to set max memlock to unlimited because of security
concerns, so I concentrated on having slurm negate this limit via its
There was little success with this, and then on reading the MPI FAQ @openmpi
for a second time, and seeing that openmpi's source code actually recurs to
libc's getrlimit() to ascertain the memlock limit, I decided this setting
was not-relevant for me. This is hunch as yet, it would be nice to get it
verified, though it won't help me resolve the problem really.
Now I'm faced with having to ask about ulimit on a mailing which is not
about ulimit, but I'm hoping for some people's experience here:
"ulimit -l unlimited" works if:
1) the /etc/security/limits.conf file has unlimited soft and hard limits.
2) if you are root.
These conditions solve the problem. But the 2nd is not feasible as should be
clear. The problem is that for a normal user running ulimit does not allow a
memlock max setting above 256. Its behaviour is one of:
"Let normaluser increase max memlock only up to 256"
Can anybody shed any light or make any suggestions, it would be appreciated.