Eygene Ryabinkin wrote:
> Arnau, good day.
>
> Thu, Nov 04, 2010 at 04:07:54PM +0100, Arnau Bria wrote:
>>> , I am running Torque 2.5.2
>> did you find any issue when running this torque version and gLite-WN 3.2 /
>> CREAMCE 3.2 / lcg-CE 3.1?
>
> Only the one with CREAM CE multiple stagein/stageout directives,
> https://savannah.cern.ch/bugs/?70808
>
>> did you see any job with negative values in its walltime?
>
> Haven't seen them recently, but I am not closely monitoring APEL
> database in this regard.
Hi,
AFAIK CERN is using oom_kill on lxplus to kill processes on lxplus
when the system reaches a high level of memory occupancy.
We saw at IFIC WNs some processes requiring 6 GBytes of VM and 2.5 of
RSS. I think it could be a good idea to use oom_kill sice it allows
running a big mem process when the total memory does not reach the
limit, but it kills some (the biggest one) when there are problems.
I will give it a try, but I haven't found the RPMS and I don't
know the license status of those. Does anybody know about this ?
Has anybody tried oom_kill on WNs ?
Regards,
Javier
--
-----------------------------------------------------------------------
| Javier Sanchez | Tel: (+34) 96.354.36.97 |
| IFIC (Instituto de Fisica Corpuscular)| Fax: (+34) 96.354.37.42 |
| CSIC-Universidad de Valencia |E-Mail: |
| Edificio Institutos de Investigación | [log in to unmask] |
| Apartado de Correos 22085 |WWW: |
| E-46071 Valencia - SPAIN | http://ific.uv.es/~sanchezj |
----------------------------------------------------------------------
|