-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi list,
I've noticed in our two clusters, that the maxjob policy behaves differently. We're running the same version and configuration, but one site limits a user with usercfg, and the other is balancing several queues with classcfg. At IU: USERCFG[user1] MAXJOB=28,200 This works as intended, when there are no other queued jobs the scheduler starts running user1 jobs. At UC: CLASSCFG[uct2] MAXJOB=8,800 CLASSCFG[prod] MAXJOB=752,936 CLASSCFG[analy_short] MAXJOB=200,952 This does not work as intended. Even when there are only prod jobs available, it limits the running jobs to 752, and leaves us with idle cpus. diagnose -q says that the queued jobs cannot run because of the soft limit of 752. What's going on here? Does Classcfg not support hard and soft limits? Has anyone gotten this or a similar configuration to work? - --Sarah - -- Sarah Williams [EMAIL PROTECTED] (317) 274-0595 http://pgp.mit.edu:11371/pks/lookup?op=get&search=0x3E6F7BBA -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.7 (Darwin) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iD8DBQFIihNH2j+lnz5ve7oRApmhAKCWlxnHvyyA4ZJWdbca1hrW4ysikACfbzkh 4IOhMUEb3wuPEDdfqvVl9us= =R1n6 -----END PGP SIGNATURE----- _______________________________________________ mauiusers mailing list [email protected] http://www.supercluster.org/mailman/listinfo/mauiusers
