[JLSE-Intel-NDA] immediate job exit on KNL nodes.
Hi JLSE support, I'm not sure if this a known issue or not. I noticed some strange behaviour last night on the cobalt for KNL. My jobs submitted to knl_7210 exited without starting. I also noticed that I can ssh directly to several nodes without being allocated by cobalt for running jobs. The cobalt log and my job script is attached. Ye log: Jobid: 548 qsub --mode script -q knl_7210 -n 4 -t 120 -O FeSe --attrs knl_memory_mode=Cache:knl_cluster_mode=Quadrant qmcpack.sub Wed Aug 17 02:48:30 2016 +0000 (UTC) submitted with cwd set to: /home/yeluo/SeTi-AB-collinear/dis_8.5/optimization-S1 jobid 548 submitted from terminal /dev/pts/9 Wed Aug 17 02:49:02 2016 +0000 (UTC) Info: resources released; starting job epilogue scripts Jobid: 549 qsub --mode script -q knl_7210 -n 1 -t 120 -O FeSe --attrs knl_memory_mode=Cache:knl_cluster_mode=Quadrant qmcpack.sub Wed Aug 17 02:49:45 2016 +0000 (UTC) submitted with cwd set to: /home/yeluo/SeTi-AB-collinear/dis_8.5/optimization-S1 jobid 549 submitted from terminal /dev/pts/9 Wed Aug 17 02:50:03 2016 +0000 (UTC) Info: resources released; starting job epilogue scripts Jobid: 550 qsub --mode script -q knl_7210 -n 1 -t 120 -O FeSe --attrs knl_memory_mode=Cache:knl_cluster_mode=Quadrant qmcpack.sub Wed Aug 17 02:53:31 2016 +0000 (UTC) submitted with cwd set to: /home/yeluo/SeTi-AB-collinear/dis_8.5/optimization-S1 jobid 550 submitted from terminal /dev/pts/9 Wed Aug 17 02:53:44 2016 +0000 (UTC) Info: resources released; starting job epilogue scripts Jobid: 551 qsub -q knl_7210 -n 1 -t 120 -O FeSe --attrs knl_memory_mode=Cache:knl_cluster_mode=Quadrant -I Wed Aug 17 02:54:08 2016 +0000 (UTC) submitted with cwd set to: /home/yeluo/SeTi-AB-collinear/dis_8.5/optimization-S1 jobid 551 submitted from terminal /dev/pts/9 Wed Aug 17 02:54:35 2016 +0000 (UTC) Info: resources released; starting job epilogue scripts script: #COBALT -q knl_7210 #COBALT -n 2 #COBALT -t 120 #COBALT -O vmc #COBALT --attrs knl_memory_mode=Cache:knl_cluster_mode=Quadrant =================== Ye Luo, Ph.D. Leadership Computing Facility Argonne National Laboratory
participants (1)
-
Luo, Ye