#!/bin/sh # # Sample job command template file used to generate cluster rolling update # jobs that will be submitted to LoadLeveler. # xCAT will substitute the following when creating the LL job command files: # [[NODESET]] - nodeset name # [[XNODELIST]] - comma-delimited list of xcat node names # [[XCATSERVER]] - xcat server name # [[XCATPORT]]] - xcatd port on server # [[LLMACHINES]] - the LL machine list. Quoted, space delimited list of LL # machine names that were available at the time the xCAT # rollupdate command was run # [[LLCOUNT]] - number of machines in LLMACHINES list # # @ job_name = rollupdate_[[NODESET]] # @ job_type = parallel ## Note: really want node usage to be not shared, but there ## was a timing bug in LL that sometimes caused a rollupdate job to ## not start. Should be fixed by now. # @ node_usage = not_shared # @ restart = no # @ error = /home/lluser/rollupdate_jobs/rollupdate_[[NODESET]].$(Host).$(Cluster).$(Process).err # @ output = /home/lluser/rollupdate_jobs/rollupdate_[[NODESET]].$(Host).$(Cluster).$(Process).out # @ executable = /opt/xcat/share/xcat/rollupdate/send_reboot_request # @ arguments = 10 --verbose loadleveler [[XCATSERVER]] [[XCATPORT]] [[XNODELIST]] # @ node = [[LLCOUNT]] # @ tasks_per_node = 1 # @ requirements = (Machine == {[[LLMACHINES]]}) # @ wall_clock_limit = 21:00,20:00 # @ queue