On Thu, May 31, 2012 at 11:55 AM, Mark Dixon <m.c.di...@leeds.ac.uk> wrote: > Please correct me if I've misunderstood but, from previous emails, I thought > you were already using these in your new cgroup code: h_vmem would now set > both memory.memsw.limit_in_bytes and setrlimit's RLIMIT_AS. > > The problem I have with this, is that some jobs benefit from being able to > specify these two quantities independently.
Hi Mark, Correct me if I am wrong, I thought your main concern was related to setting address space limit with memsw. The biggest difference is that with RLIMIT, malloc would return NULL vs in the memsw case the job would be killed by the kernel OOM killer, and that's not the behavior you want. So what I've asked Ron to do is to check with the kernel guys and see if they can provide us the semantics of setrlimit() with memory.memsw.limit_in_bytes, or add a new limit in the memory cgroup controller. (May be I should have clarified the above point in my previous email - but I was really busy these days, working on the GE2011.11u1 release, handling outside of the mailing list user support, and talking to hardware vendors, etc...) Rayson > > As I don't want to go down one route with my patchset while the rest of the > community goes down another, my question is: > > Do you support the idea of introducing a new gridengine attribute to control > memory.memsw.limit_in_bytes and leave h_vmem to just deal with setrlimit's > RLIMIT_AS, as previously? > > > Best wishes, > > Mark > -- > ----------------------------------------------------------------- > Mark Dixon Email : m.c.di...@leeds.ac.uk > HPC/Grid Systems Support Tel (int): 35429 > Information Systems Services Tel (ext): +44(0)113 343 5429 > University of Leeds, LS2 9JT, UK > ----------------------------------------------------------------- _______________________________________________ users mailing list users@gridengine.org https://gridengine.org/mailman/listinfo/users
