Am 22.08.2012 um 11:35 schrieb William Hay: > I'm looking to add support for BLCR into our grid engine > configuration. A quick google suggests the scripts here > https://github.com/HPCKP/BLCR-GridEngine-Integration should be useful > for such integration. Not wishing to engage in cargo cult sysadmin > I'm trying to understand what the scripts actually do. It appears > that the id to be fed to cr_checkpoint is supposed to be generated by > the following command: > pstree -p $pid | head -1 | perl -pe '$p="g\?time"; $p=cr_restart > if(/cr_restart\(\d+\)/);s/.*-$p\(\d+\)[-\+]+[^(]+\((\d+)\)/$1/g;' > > As far as I can tell this attempts to extract from the first line of > pstree -p $pid's output the process id of the first child of > cr_restart command if present or time/gtime if not. cr_restart should
> cr_restart command if present or of the time/gtime if not. cr_restart should I'm not used to BLCR, but there is an awk explanation above in the script. Does this work better? This is similar to the one in the http://arc.liv.ac.uk/SGE/howto/APSTC-TB-2004-005.pdf where the last PID in the line is output. -- Reuti > presumably be the ancestor of the useful parts of jobs that have been > restored at least once but I can't see any reason in the scripts to > expect that time will be the ancestor of all the useful parts of a job > prior to the first restart. Feeding pids that > don't have time or cr_restart as a descendant on the first line just > produces the first line of pstree -p's output which doesn't look like > it would be useful for feeding to cr_chckpoint. > > William > _______________________________________________ > users mailing list > users@gridengine.org > https://gridengine.org/mailman/listinfo/users _______________________________________________ users mailing list users@gridengine.org https://gridengine.org/mailman/listinfo/users
