[gridengine users] Memory limits in SGE and limits.conf
reuti at staff.uni-marburg.de
Fri Mar 18 12:47:49 UTC 2011
Am 18.03.2011 um 12:39 schrieb Arne Brutschy:
> I guess this has been covered before, but I wasn't able to find an
> answer in the list archives (so, sorry if this is kind-of-a-faq)
> We're limiting memory on our cluster using /etc/security/limits.conf.
> Somehow, this does not apply to submitted jobs, as we've just found
> out. I guess the proper way to limit mem in GE is to use s_vmem and
> h_vmem - but why does the limits.conf method not work?
this is correct. A job started by `qsub` is a fork of the execd, which switches at one point to the target user. The /etc/security/limits.conf is checked by PAM which is not in SGE's current implementation to be accessed during job startup.
All settings of limits like user_lists, xuser_lists, h_vmem, h_stack,... are therefore in SGE and the same for all platforms you run on (even in heterogenous clusters mixing Solaris, Irix and Linux). In fact: it might be confusing having too many places where a limit can be specified. There is already the option to specify it also in an RQS in SGE, to limit it per user in the complete cluster for example.
> Arne Brutschy
> Ph.D. Student Email arne.brutschy(AT)ulb.ac.be
> IRIDIA CP 194/6 Web iridia.ulb.ac.be/~abrutschy
> Universite' Libre de Bruxelles Tel +32 2 650 2273
> Avenue Franklin Roosevelt 50 Fax +32 2 650 2715
> 1050 Bruxelles, Belgium (Tel and Fax both IRIDIA secretary)
> users mailing list
> users at gridengine.org
More information about the users