On Jul 24, 2013, at 10:01 AM, Romain <nuelromain@xxxxxxxxx> wrote: > Romain <nuelromain@...> writes: > >> >> Hi everybody, >> >> I would like to know how to configure my execute nodes to preempt jobs > that >> exceed the memory reserved for HTCondor. >> >> I use Ubuntu 12.04 on execute nodes and I have HTCondor version 7.8.7 >> >> I look at https://htcondor-wiki.cs.wisc.edu/index.cgi/wiki? >> p=HowToLimitMemoryUsage >> and in the documentation but my tests are not concluding. >> >> I would preempt jobs when they exceed the limit so I do this: >> PREEMPT = $(PREEMPT) || ((MemoryUsage > Memory) =!= TRUE) >> WANT_SUSPEND = $(WANT_SUSPEND) && ((MemoryUsage > Memory) =!= TRUE) >> (like in the wiki) >> >> But that doesn't work, and I try to replace "Memory" by a number (1000 for >> example and that doesn't work too. >> >> I try USER_JOB_WRAPPER but I don't really understand how to use it... >> >> Thank you. >> Have a nice day. >> >> -- >> Romain >> >> > > Ok so I try use cgroups. > > I've install it and create a group for htcondor like in the documentation > but what I have to add in configuration file or submit file ? because is not > clear in documentaition. > > I just add this in configuration file: > BASE_CGROUP = htcondor > CGROUP_MEMORY_LIMIT_POLICY = hard > > And I set up the /htcondor/memory.limit_in_bytes to 500MB but my test are > not concluding > > Thanks. > Did you add that in the worker node's configuration file? How are you testing - and what makes you think it isn't working? Brian
Attachment:
smime.p7s
Description: S/MIME cryptographic signature