Hi all,
Öttl Peter wrote:
> we run glite-TORQUE_server-3.2.2-0 which uses torque-2.3.6-2. The 2.3 branch of
> torque is due one more revision before it becomes end of life (which will be towards the end
> of September/October)
>
> Does anybody know why the gLite repo is not moving forward to the 2.4 branch?
>
>
The main reason is because nobody is officially responsible for creating
new versions of torque ... At Nikhef we agreed to support torque on a
best-effort basis , but we would not produce new versions of torque:
we'd use whatever is in the EPEL repositories. The current version in
EPEL is 2.3.10, uploaded by Steve Traylen. If nobody updates the EPEL
repos then no major new version would be introduced.
I've brought this issue up within EMI and even within EGI but there's no
consensus on how to proceed.
My main questions are:
- what are the requirements for a new version of Torque/PBS? which PTs
are affected?
- if we upgrade, why not upgrade to 2.5.2 instead of 2.4.10?
- who will maintain the EPEL version of torque (currently Steve Traylen
is listed but I know he has many other things to do)?
Similar questions apply to the other batch systems, BTW: what about Sun
Grid Engine? How are the SGE maintainers informed of changes in the
interfaces between the batch system and glite? what is there's a major
change in the blah scripts - how are the different batch systems
affected? who "owns" this interface ? Matters like this were covered in
EGEE III by the SA3 batch system integration team (Dennis and me) but
this has all disappeared in the post-EGEE III era.
> I'm pretty sure that some sites already tested the 2.4 branch or have it in production.
> It would be nice if we could share experiences and move forward with the torque version for the gLite repo.
>
> We installed 2.4.10-1 the latest stable release on a test cluster recently and up to now job submission and
> the information system look fine. We did not test apel-pbs-log-parser though.
>
> Does somebody have a list of services that have to be tested against torque?
the list of services to test would be
- cream-ce : blah scripts and gip plugin
- APEL parser (which we do not use either at Nikhef)
- lcg-ce (if required)
> On Sep 15, 2010, at 11:46 AM, Massimo Sgaravatto - INFN Padova wrote:
>
>
>> Yes, it was reported to the BLAH developers:
>>
>> http://savannah.cern.ch/bugs/?70808
>>
>> who agreed to implement the proposed patch.
>>
>> But, just to be fully synchronized (TM), as far as I can understand
>> (please correct me if I am wrong), the bug is in Torque !
>>
>> Aren't they (torque developers) going to fix the problem ?
>>
>
> I found this in the changelog of torque 2.4.9:
>
> b - restore ability to parse -W x=geometry:{...,...}
>
> @Tim: Did you test this with 2.4.9 or with an earlier release?
>
>
see
http://www.clusterresources.com/bugzilla/show_bug.cgi?id=80
this bug was reported (again) to the torque developers one week ago but
they're still in denial...
There's also a patch from Eygene Ryabinkin to solve this issue for both
2.4.9+ and 2.5 .
I'd propose that IF we upgrade that
- we upgrade to 2.5.2
- we include Eygene's patch to solve the -W issue
share and enjoy,
JJK / Jan Just Keijser
Nikhef
Amsterdam
|