2574 2005-01-18 10:24 fork job not being killed at resource destruction 2005-04-01 12:00:31 1 1 1 Unclassified GRAM wsrf managed execution job service 3.9.4 PC Linux RESOLVED FIXED P1 blocker 4.0 1 tfreeman@mcs.anl.gov smartin@mcs.anl.gov alain@isi.edu bester@mcs.anl.gov gaffaney@mcs.anl.gov lane@mcs.anl.gov madduri@mcs.anl.gov rynge@isi.edu smartin@mcs.anl.gov tfreeman@mcs.anl.gov 2005-01-18 10:24:11 When ManagedExecutableJobResource.remove() is called, no signals are being sent to the forked job pid. On early cancel, the job still continues. I also put a touch file in the fork scheduler adapter cancellation function which does not get touched, so I think the resource->sudo->perl link is not there for destroy. lane@mcs.anl.gov 2005-01-21 17:44:41 Fixed in trunk. lane@mcs.anl.gov 2005-02-01 11:32:41 *** Bug 2672 has been marked as a duplicate of this bug. *** lane@mcs.anl.gov 2005-02-01 11:33:52 Bob, how recent is your GRAM installation? As noted in comment #1, a fix was committed on the 1/21/ 2005. gaffaney@mcs.anl.gov 2005-02-01 13:46:50 I tried this again for both Condor and PBS with small numbers of jobs. For my installation (ws gram rebuilt today) the jobs remained enqueued with the scheduler after globusrun-ws -kill had been run. ***** Log for a manual submission of PBS job below globusrun-ws -submit -factory https://lucky0:8444/wsrf/services/ManagedJobFactoryService -factory-type PBS - batch -o epr_0 -c /bin/sleep 1000 Submitting job...Done. Job ID: uuid:d12b8280-7488-11d9-a2b0-0002a5ad41e5 Termination time: 02/02/2005 19:38 GMT Done [gaffaney@lucky0 one]$ qstat Job id Name User Time Use S Queue ---------------- ---------------- ---------------- -------- - ----- 3048.lucky0 STDIN gaffaney 0 R luckyq [gaffaney@lucky0 one]$ globusrun-ws -kill -j epr_0 Requesting original job description...Done. Destroying job...Done. [gaffaney@lucky0 one]$ qstat Job id Name User Time Use S Queue ---------------- ---------------- ---------------- -------- - ----- 3048.lucky0 STDIN gaffaney 0 E luckyq [gaffaney@lucky0 one]$ qstat Job id Name User Time Use S Queue ---------------- ---------------- ---------------- -------- - ----- 3048.lucky0 STDIN gaffaney 0 E luckyq gaffaney@mcs.anl.gov 2005-02-01 13:49:54 I looked at what I pasted again and see that the state of the enqueued job went from R to E - This may be ok. gaffaney@mcs.anl.gov 2005-02-01 13:53:37 OK - it might work for PBS but Condor still seems to be a problem - paste below: ***************** globusrun-ws -submit -factory https://lucky0:8444/wsrf/services/ManagedJobFactoryService -factory-type Condor -batch -o epr_0 -c /bin/sleep 1000 Submitting job...Done. Job ID: uuid:84c5203e-748a-11d9-a422-0002a5ad41e5 Termination time: 02/02/2005 19:50 GMT Done [gaffaney@lucky0 one]$ condor_q -- Submitter: lucky0.mcs.anl.gov : <140.221.65.193:58106> : lucky0.mcs.anl.gov ID OWNER SUBMITTED RUN_TIME ST PRI SIZE CMD 3011.0 gaffaney 2/1 13:50 0+00:00:01 R 0 0.0 sleep 1000 1 jobs; 0 idle, 1 running, 0 held [gaffaney@lucky0 one]$ globusrun-ws -kill -j epr_0 Requesting original job description...Done. Destroying job... Done [gaffaney@lucky0 one]$ condor_q -- Submitter: lucky0.mcs.anl.gov : <140.221.65.193:58106> : lucky0.mcs.anl.gov ID OWNER SUBMITTED RUN_TIME ST PRI SIZE CMD 3011.0 gaffaney 2/1 13:50 0+00:00:33 R 0 0.0 sleep 1000 1 jobs; 0 idle, 1 running, 0 held [gaffaney@lucky0 one]$ condor_q -- Submitter: lucky0.mcs.anl.gov : <140.221.65.193:58106> : lucky0.mcs.anl.gov ID OWNER SUBMITTED RUN_TIME ST PRI SIZE CMD 3011.0 gaffaney 2/1 13:50 0+00:00:46 R 0 0.0 sleep 1000 1 jobs; 0 idle, 1 running, 0 held [gaffaney@lucky0 one]$ condor_q -- Submitter: lucky0.mcs.anl.gov : <140.221.65.193:58106> : lucky0.mcs.anl.gov ID OWNER SUBMITTED RUN_TIME ST PRI SIZE CMD 3011.0 gaffaney 2/1 13:50 0+00:00:49 R 0 0.0 sleep 1000 1 jobs; 0 idle, 1 running, 0 held [gaffaney@lucky0 one]$ lane@mcs.anl.gov 2005-02-02 12:24:00 If it's working for other schedulers, then this is an issue with either Condor or the perl scheduler script for Condor. Reassigning to Stu so he can determine who should be looking at this. lane@mcs.anl.gov 2005-02-16 11:42:38 *** Bug 2749 has been marked as a duplicate of this bug. *** lane@mcs.anl.gov 2005-02-16 11:50:55 Ignore the duplicate assignment. It was intended for #2575. lane@mcs.anl.gov 2005-04-01 12:00:31 I'm closing this since Fork and PBS job removal is working. Condor has a separate bug open for it.