I am running Maui version 3.2.6p19 on a ROCKS 4 cluster.
All nodes have 32GB of RAM and a 32GB swap partition for a total of
64GB virtual memory.
When I run 'checknode' on a node, the 'Configured Resources' lines
always gives static 'PROCS: 8', 'MEM: 31G' and 'DISK: 1M' (no
idea what DISK is). However SWAP is always variable and always
some amount less than 64G. For example:
========================================================================
# checknode compute-0-61
checking node compute-0-61
State: Running (in current state for 00:00:00)
Configured Resources: PROCS: 8 MEM: 31G SWAP: 48G DISK: 1M
Utilized Resources: [NONE]
Dedicated Resources: PROCS: 3 SWAP: 48G
Opsys: linux Arch: [NONE]
Speed: 1.00 Load: 3.000
Network: [DEFAULT]
Features: [nonGPU][rack12]
Attributes: [Batch]
Classes: [p20 8:8][p30 8:8][p40 8:8][p5 8:8][p60 8:8][matlab 8:8][default
5:8][GPU 8:8][extended 8:8][p10 8:8]
Total Time: INFINITY Up: INFINITY (84.84%) Active: INFINITY (39.53%)
Reservations:
Job '1410460'(x1) -2:00:58:34 -> 1:23:01:26 (4:00:00:00)
Job '1410461'(x1) -2:00:58:28 -> 1:23:01:32 (4:00:00:00)
Job '1415119'(x1) -5:55:33 -> 3:18:04:27 (4:00:00:00)
JobList: 1410460,1410461,1415119
========================================================================
On others nodes for example I see
Configured Resources: PROCS: 8 MEM: 31G SWAP: 59G DISK: 1M
Configured Resources: PROCS: 8 MEM: 31G SWAP: 51G DISK: 1M
Configured Resources: PROCS: 8 MEM: 31G SWAP: 59G DISK: 1M
Configured Resources: PROCS: 8 MEM: 31G SWAP: 58G DISK: 1M
Configured Resources: PROCS: 8 MEM: 31G SWAP: 59G DISK: 1M
Configured Resources: PROCS: 8 MEM: 31G SWAP: 51G DISK: 1M
I don't think this is just a display bug but affects scheduling too.
In the first example above there are 3 jobs that were submitted
with vmem=20gb, vmem=20gb and vmem=8gb respectively giving the 48G
seen on the 'Dedicated' line.
Since 48G on the 'Dedicated' line is >= the 48G on the 'Configured'
line, maui appears to be refusing to run any more jobs on this node
even though there really is free SWAP.
On this node, if I run 'free' I see:
total used free shared buffers cached
Mem: 32962780 16729048 16233732 0 335652 1782692
-/+ buffers/cache: 14610704 18352076
Swap: 32764556 0 32764556
so that 48G on the 'Configured' line seems to be a calculation of
total virtual memory minus actual used real memory. But that seems
to be the wrong thing to do.
Is there anyway in configuration to just force SWAP to be 64G for all
nodes?
---------------------------------------------------------------
Paul Raines email: raines at nmr.mgh.harvard.edu
MGH/MIT/HMS Athinoula A. Martinos Center for Biomedical Imaging
149 (2301) 13th Street Charlestown, MA 02129 USA
The information in this e-mail is intended only for the person to whom it is
addressed. If you believe this e-mail was sent to you in error and the e-mail
contains patient information, please contact the Partners Compliance HelpLine at
http://www.partners.org/complianceline . If the e-mail was sent to you in error
but does not contain patient information, please contact the sender and properly
dispose of the e-mail.
_______________________________________________
mauiusers mailing list
[email protected]
http://www.supercluster.org/mailman/listinfo/mauiusers
