Hello,
our local cluster is running less jobs than it actually could; digging
into that, we found that the value that MAUI labels as "SWAP" is
probably incorrect.
Indeed::
$ checknode wn03
[...]
State: Busy (in current state for 00:13:30)
Configured Resources: PROCS: 16 MEM: 31G SWAP: 14G DISK: 1M
^^^^^^^^^
Utilized Resources: PROCS: 16
Dedicated Resources: PROCS: 15 MEM: 29G SWAP: 12G
[...]
But the node has only 10G of swap, of which 8G are free::
$ ssh wn03 free -m
total used free shared buffers cached
Mem: 32484 32316 168 0 26 5989
-/+ buffers/cache: 26301 6183
Swap: 10236 2239 7996
It looks like MAUI takes its "SWAP" value by looking at TORQUE's "availmem"::
$ pbsnodes -a
wn03.lcg.cscs.ch
state = free
[...]
status =
[...],totmem=43746664kb,availmem=14381788kb,physmem=33264260kb,[...]
^^^^^^^^^^^^^^^^^^^
Shouldn't "SWAP" reflect either the total node memory or the size of
the swap partition?
We're using MAUI 3.2.6p20 on SLC4 as found in the gLite distribution::
$ rpm -qa | fgrep maui
maui-3.2.6p20-snap.1182974819.8.slc4
maui-client-3.2.6p20-snap.1182974819.8.slc4
maui-server-3.2.6p20-snap.1182974819.8.slc4
Best regards,
Riccardo
--
Riccardo Murri
CSCS - Swiss National Centre for Supercomputing
Galleria 2, via Cantonale
CH-6928 Manno (Switzerland)
tel.: +41 91 610 8234
Fax: +41 91 610 8282
_______________________________________________
mauiusers mailing list
[email protected]
http://www.supercluster.org/mailman/listinfo/mauiusers
