You are here: 21 Preemption > How-to > Using Owner Preemption

21.3 Using Owner Preemption

To enable owner preemption

  1. Make the following configurations to the moab.cfg file:
    1. Set GUARANTEEDPREEMPTION to TRUE. (This causes Moab to lock PREEMPTOR jobs until JOBRETRYTIME expires.)
    2. Make sure that JOBNODEMATCHPOLICY is not set to EXACTNODE, which is not currently supported for preemption (for more information, see Testing and Troubleshooting Preemption).
    3. Set the PREEMPTPOLICY type (for more information, see PREEMPTPOLICY Types).
    4. Set the OWNERPREEMPT flag.
    5. Optionally, if you want the owner preemption to override any PREEMPTMINTIME settings for PREEMPTEE jobs, you can set the OWNERPREEMPTIGNOREMINTIME flag as well.

    6. Specify an owner.
    7. If the non-owner job does not have a RESTARTABLE or REQUEUEABLE flag set, the job will cancel.

  2. For example:

    GUARANTEEDPREEMPTION TRUE
    PREEMPTPOLICY <policy>
     
    SRCFG[myrez] FLAGS=OWNERPREEMPT HOSTLIST=node01
    SRCFG[myrez] OWNER=USER:john
    SRCFG[myrez] USERLIST=jane,john PERIOD=INFINITY
     
    QOSCFG[test1] QFLAGS=PREEMPTEE JOBFLAGS=restartable MEMBERULIST=john PRIORITY=100
    QOSCFG[test2] QFLAGS=PREEMPTOR MEMBERULIST=john PRIORITY=10000
  3. Submit a job to a user who is not the owner (in this example, jane).
  4. [jane@g06]$ echo sleep 600 | msub -l walltime=600 -l procs=64

    (Optional) Examine the following output for showq and checkjob for jane's job:

    Moab.1
    [jane@g06]$ showq
     
    active jobs------------------------
    JOBID     USERNAME    STATE      PROCS      REMAINING     STARTTIME
    Moab.1    jane        Running    64         00:09:57      Mon Nov 14 12:07:52
     
    1 active job     64 of 64 processors in use by local jobs (100.00%)
                     1 of 1 nodes active (100.00%)
     
    eligible jobs----------------------
    JOBID     USERNAME    STATE      PROCS      WCLIMIT       QUEUETIME
     
    0 eligible jobs
     
    blocked jobs-----------------------
    JOBID     USERNAME    STATE      PROCS      WCLIMIT       QUEUETIME
     
    0 blocked jobs
     
    Total job: 1
    root@g06]# checkjob Moab.1
    job Moab.1
     
    State: Running
    Creds: user:jane group:jane
    WallTime: 00:01:02 of 00:10:00
    SubmitTime: Mon Nov 14 12:07:52
    (Time Queued Total: 00:00:00 Eligible: 00:00:00)
     
    StartTime: Mon Nov 14 12:07:52
    Total Requested Tasks: 64
     
    Req[0] TaskCount: 64 Partition: FLEXlm
    NodeCount: 1
     
    Allocated Nodes:
    [node01:64]
     
     
    IWD: /opt/native
    SubmitDir: /opt/native
    Executable: /opt/native/spool/moab.job.FoZfIU
     
    StartCount: 1
    Flags: GLOBALQUEUE,PROCSPECIFIED
    StartPriority: 1
    Reservation 'Moab.1' (-00:01:24 -> 00:08:36 Duration: 00:10:00)
  5. Now submit a job for the owner (in this example, john).
  6. [john@g06]$ echo sleep 600 | msub -l walltime=600 -l procs=50

    [john@g06]$ echo sleep 600 | msub -l walltime=600 -l procs=50

    (Optional) Examine the following output for showq and checkjob for john's job:

    Moab.2
    [john@g06]$ showq
     
    active jobs------------------------
    JOBID     USERNAME    STATE      PROCS     REMAINING     STARTTIME
    Moab.1    jane        Canceling  64        00:07:43      Mon Nov 14 12:07:52
    Moab.2    john        Running    50        00:09:59      Mon Nov 14 12:10:08
     
    2 active jobs    64 of 64 processors in use by local jobs (100.00%)
                     1 of 1 nodes active (100.00%)
     
    eligible jobs----------------------
    JOBID     USERNAME    STATE     PROCS      WCLIMIT       QUEUETIME
     
    0 eligible jobs
     
    blocked jobs-----------------------
    JOBID     USERNAME    STATE     PROCS      WCLIMIT       QUEUETIME
     
    0 blocked jobs
     
    Total jobs: 2

    Note that jane's job is canceled once john's job is submitted.

    [john@g06]$ checkjob Moab.2
    job Moab.2
     
    State: Running
    Creds: user:john group:john
    WallTime: 00:00:31 of 00:10:00
    SubmitTime: Mon Nov 14 12:10:08
    (Time Queued Total: 00:00:00 Eligible: 00:00:00)
     
    StartTime: Mon Nov 14 12:10:08
    Total Requested Tasks: 50
     
    Req[0] TaskCount: 50 Partition: FLEXlm
    NodeCount: 1
     
    Allocated Nodes:
    [node01:50]
     
     
    IWD: /opt/native
    SubmitDir: /opt/native
    Executable: /opt/native/spool/moab.job.jf1N4a
     
    StartCount: 1
    Flags: HASPREEMPTED,GLOBALQUEUE,PROCSPECIFIED
    StartPriority: 1
    Reservation 'Moab.2' (-00:00:48 -> 00:09:12 Duration: 00:10:00)

    Note the new HASPREEMPTED flag.

    (Optional) Now look at the showq for jane's job (after):

    [root@g06]# checkjob Moab.1
    job Moab.1
     
    State: Removed
    Completion Code: -1 Time: Mon Nov 14 12:10:08
    Creds: user:jane group:jane
    WallTime: 00:02:47 of 00:10:00
    SubmitTime: Mon Nov 14 12:07:52
    (Time Queued Total: 00:00:00 Eligible: 00:00:00)
     
    Total Requested Tasks: 64
     
    Req[0] TaskCount: 64 Partition: FLEXlm
    NodeCount: 1
     
    Allocated Nodes:
    [node01:64]
     
     
    IWD: /opt/native
    Executable: /opt/native/spool/moab.job.FoZfIU
     
    Execution Partition: FLEXlm
    Flags: GLOBALQUEUE,PROCSPECIFIED
    StartPriority: 0

    Note that the state is now Removed.

Related Topics 

© 2016 Adaptive Computing