Hi Loris, ok, THAT seems really much. What do you use for gathering these values? jobacct_gather/cgroup?
If I remember right, there was a discussion lately in this list regarding the JobAcctGatherType, yet I do not remember the outcame. I remember though, that someone pointed to SLUG18 (or 17?). Nonetheless there is the following whitepaper:
https://slurm.schedmd.com/SLUG18/field_notes2.pdf My current recommended deployment has: ○ proctrack/cgroup, task/cgroup, jobacct_gather/cgroup ○ cgroup.conf - ConstrainCores, ContrainDevices, ConstrainRAMSpace, ConstrainSwapSpace all enabled ○ PrologFlags=contain with pam_slurm_adopt setup appropriately ○ LaunchParameters=send_gids ○ ReconfigFlags=KeepPartInfo Best Marcus On 2/26/19 1:20 PM, Loris Bennett wrote:
Hi Marcus, Thanks for the response, but that doesn't seem to be the issue. The problem seems to be that the raw data are incorrect: Slurm data: ... Ncpus Nnodes Ntasks Reqmem PerNode Cput Walltime Mem ExitStatus Slurm data: ... 50 2 1 102400000 0 503611 16310 1.8014398509482e+16 0 Cheers, Loris Marcus Wagner <[email protected]> writes:Hi Loris, I assume, this job used FAIRLY few memory, in the kb range, might that be true? replace sub kbytes2str { my $kbytes = shift; if ($kbytes == 0) { return sprintf("%.2f %sB", 0.0, 'M'); } my $mul = 1024; my $exp = int(log($kbytes) / log($mul)); my @pre = qw/ M G T P E /; my $pre = $pre[$exp-1]; return sprintf("%.2f %sB", ($kbytes / pow($mul, $exp)), $pre ? $pre : ""); } with my @pre = qw/ k M G T P E /; my $pre = $pre[$exp]; Best Marcus On 2/26/19 10:08 AM, Loris Bennett wrote:48.00 EB (estimated maximum) Memory Efficiency: 26388279066.62% of 195.31 GB (1.95 GB/core)
-- Marcus Wagner, Dipl.-Inf. IT Center Abteilung: Systeme und Betrieb RWTH Aachen University Seffenter Weg 23 52074 Aachen Tel: +49 241 80-24383 Fax: +49 241 80-624383 [email protected] www.itc.rwth-aachen.de
