- 09 Nov, 2012 8 commits
-
-
Morris Jette authored
-
Danny Auble authored
Conflicts: src/srun/libsrun/opt.c
-
Danny Auble authored
-
Danny Auble authored
and perhaps someone might find it useful).
-
Danny Auble authored
-
Morris Jette authored
-
Danny Auble authored
-
Danny Auble authored
-
- 08 Nov, 2012 16 commits
-
-
Danny Auble authored
-
Morris Jette authored
-
Danny Auble authored
One change is a check was added to see if Limits are enforced before doing any of the work. Before it wasn't part of the mix so some jobs could get killed if limits were enforced and then turned off.
-
Danny Auble authored
enforcing a QOS doesn't have anything to do with enforcing a limit if a QOS is used.
-
Danny Auble authored
-
Danny Auble authored
Signed-off-by:
Danny Auble <da@schedmd.com>
-
Danny Auble authored
-
Morris Jette authored
-
Janne Blomqvist authored
running the perf profiler on slurmctld on our production cluster (with select/cons_res) showed that with our configuration & workload highest up in the profile are calls to difftime() and _cr_job_list_sort(), in total accounting for ~37% of the total runtime. By sacrificing some theoretical portability by replacing the calls to difftime() with a simple inline subtraction that 37% should be substantially reduced. As I saw identical code also in the linear and serial plugins, I fixed those in the same manner, although I don't know if this performance issue turns up when using those plugins.
-
Morris Jette authored
-
Danny Auble authored
-
Danny Auble authored
-
Danny Auble authored
-
Morris Jette authored
-
Morris Jette authored
-
Morris Jette authored
-
- 07 Nov, 2012 16 commits
-
-
Danny Auble authored
Conflicts: NEWS configure doc/html/cray.shtml
-
Danny Auble authored
-
Danny Auble authored
-
Danny Auble authored
-
Danny Auble authored
-
Danny Auble authored
than BGQ.
-
Morris Jette authored
-
Morris Jette authored
-
Morris Jette authored
Remove some unused variables and minor reformatting to conform with Linux kernel coding standard
-
Danny Auble authored
aprun process instead of a perl script.
-
Morris Jette authored
Conflicts: configure doc/html/documentation.shtml src/common/cpu_frequency.c src/common/slurm_jobacct_gather.c src/common/slurm_protocol_defs.c src/common/slurm_protocol_defs.h src/plugins/accounting_storage/mysql/as_mysql_job.c src/sstat/sstat.c
-
Yiannis Georgiou authored
The patch contains: 1) general code cleanup 2) documentation concerning the energy framework 3) fix for sstat header/data mismatch problem 4) changes on the rapl plugin to print the average power consumption between 2 measurements in place of an instantaneous value: this is printed for CurrentWatts value of "scontrol show node" 5)the memory leak that Moe located on cpu frequency code we are currently working on the following : - the requested frequency as described by Martin - rapl fine grain measurements (socket level) - the ipmi plugin based on freeipmi and we hope to have them either during your 2.5.0.rc1 testing period or for the 2.5.1
-
Morris Jette authored
-
Morris Jette authored
-
Janne Blomqvist authored
we recently saw an issue where on a couple clusters the age priority of almost all jobs had been zeroed. Restarting slurmctld did "solve" the issue, though I'm not really sure why it occured in the first place. One suspicion is that it was due to the recent DST transition (occured on Oct 28th over here), and that this somehow wedged the calculation of the age priority. Looking at the code, I couldn't really find any obvious fault, but anyway, by avoiding the usage of broken-down time (which includes DST information) in the calculation of the start_time this sort of bug could at least be avoided. The attached patch implements the above for the priority multifactor plugins.
-
Janne Blomqvist authored
-