[torquedev] Resource limits not set for processes started via tm_spawn
glen.beane at gmail.com
Mon Jun 15 20:07:02 MDT 2009
On Thu, Jun 11, 2009 at 8:18 PM, Chris Samuel<csamuel at vpac.org> wrote:
> Hi all,
> Things are still pretty hectic here and I've not had time
> to follow the lists for a few weeks, sorry about that! :-(
> One of the things that has been keeping me busy is users
> who run code that pushes nodes out of memory, we're looking
> at setting default pvmem limits for everyone (initially we'd
> done that with pmem not realising that it didn't set a limit
> that would be honoured for non-trivial allocations) and I've
> been testing this out on myself.
> However, what seems to be the case is that mom_set_limits()
> is only called for the first process in the job, and not
> for any tasks started via tm_spawn. This results in pbs_mom
> killing off processes that exceed these limits, rather than
> them failing nicely when they call malloc() et. al.
> I was thinking it would be good to insert a call to
> mom_set_limits() in start_process() so that these limits
> are set as expected (well, expected by me at least :) ).
> Thoughts ?
you should put this into bugzilla!
Your suggestion seems to make sense to me, but I haven't looked into
it in much detail.
More information about the torquedev