Ok,

So I tried exchanging the sge_qmaster (6.2u5) binary for the sge_qmaster (OGS/GE 2011.11p1) that comes with rocks-6.1 and it now seems to hold its ground.

The last time I tried that and it failed, I took the binaries from the OGS site. They where the 2011.11 (no p1) and it seems that version didn't have this issue fixed.

Thanks and sorry for all this name confusion,

Txema


El 17/09/13 20:26, Luca Clementi escribió:
On Mon, Sep 16, 2013 at 3:12 PM, Dave Love <d.l...@liverpool.ac.uk> wrote:
Txema Heredia Genestar <txema.llis...@gmail.com> writes:

Hi all,

I have a cluster in production running rocks-cluster 6.0 using SGE6.2u5.
SGE6.2u5 has a bug that kills the qmaster when an amount of jobs using
both -pe and -hold_jid are used. OGE (theoretically) has this bug
fixed.
OGE is Oracle Grid Engine, which is rumoured to be trademarked.  Do you
mean that?  If so, you presumably want to talk to Oracle.

I don't know much about Rocks, but it seems to throw up a lot of
grid engine problems and doesn't ship an up-to-date SGE.

If someone can tell me how to build/install for Rocks, I'm happy to host
instructions or binaries for SGE 8.1.4 or newer, which has >1000 patches
over SGE6.2u5/OGS, including
<https://arc.liv.ac.uk/trac/SGE/changeset/3511/sge>, which is probably
the fix in question.

What is the safest/cleanest way to upgrade from SGE to OGE?
If you want to upgrade from 6.2u5 to 8.1.4, you can do it live
<https://arc.liv.ac.uk/repos/darcs/sge/source/README.upgrade> in a
vanilla system, but I don't know if that will work in Rocks.

On the "official" SGE roll (the name is there since the pre Oracle
acquisition), we have been using OGS.

http://gridscheduler.sourceforge.net/

The version released with rocks 6.1 was GE2011.11p1 which is still the
current version accordingly to the web site.

Generally before making a new Rocks release we upgrade OGS to the
latest stable release available.

Sincerely,
Luca
_______________________________________________
users mailing list
users@gridengine.org
https://gridengine.org/mailman/listinfo/users

Reply via email to