Test cluster queue setup

There are a few changes in the queuing setup:

  • On the "old" cluster, there were five queues: veryshort, short, medium, long, verylong. On the new cluster, there is only one queue: "all.q". No need to request it either: it's default.
  • You will have to request the maximum runtime, using the qsub requestable h_rt: "qsub -l h_rt=HH:MM:SS". If you don't specify a runtime, it is set to 10 minutes!
  • By default, a job gets 10 GB of memory (h_vmem). If you need more, ask for it: "-l h_vmem=100G". Memory requests are now per job, independent of the number of slots requested.
  • You can now request your own local space in /tmp, using the tmpspace requestable: qsub -l tmpspace=100G. In your job, use the "$TMPDIR" environment variable to locate it. Try to avoid using /tmp itself!
  • Unfortunately, when using qlogin in combination with the "tmpspace" requestable, the "$TMPDIR" environment variable is absent (bug...). You do have the local tempspace though; it will be located at something like "/tmp/YOUR_JOB_ID.1.default".

-- Martin Marinus - 2015-11-19


This topic: HPC > WebHome > TestCluster > TestClusterQueueSetup
Topic revision: r1 - 2015-11-19 - MartinMarinus
This site is powered by the TWiki collaboration platform Powered by PerlCopyright © 2008-2021 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback