[slurm-users] blastx fails with "Error memory mapping"

[Mahmood Naderan]

Excuse me, I have confused with that.
While the cgroup value is 68GB, I run on terminal and see the VSZ is about
80GB and the program runs normally.
However, with slurm on that node, I can not run.

Why on terminal I can run, but I can not run via slurm?
I wonder if slurm gets the right value from kernel's cgroup.


I would like to locally solve the problem for blast and I am not seeking a
system wide solution right now.

Regards,
Mahmood




On Fri, Jan 24, 2020 at 8:45 PM Mark Hahn <hahn at mcmaster.ca> wrote:

>
> of course not.  "usage_in_bytes" is an output parameter.
> your issue is that Slurm is setting at least some of the input
> controls such as memory.memsw.limit_in_bytes.  if you want to fight
> with Slurm, you could set the memory.memsw.limit_in_bytes value on
> a "live" cgroup.  (note also that above you're referring to the base
> cgroup, not the cgroup for your job.)  of course, manually fighting
> Slurm is a Fairly Bad Idea.
>
> you should read the documentation on cgroups to understand how these work.
> memsw basically corresponds to VSZ in ps, whereas mem corresponds with RSS.
>
> regards, mark hahn.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.schedmd.com/pipermail/slurm-users/attachments/20200124/25dd2f54/attachment.htm>