[gridengine users] install error
rayson at scalablelogic.com
Wed Jan 25 17:43:11 UTC 2012
On Wed, Jan 25, 2012 at 12:18 PM, Gerard Henry <ghenry at cmi.univ-mrs.fr> wrote:
> is it possible to compile sge_execd for lx24 today?
Yes, you can either:
1) build on a Linux 2.4 machine.
2) Change the "dist/util/arch" script and make it print "lx24-amd64"
instead of "lx26-amd64".
However, the easiest way *and* without recompiling or hacking is to
create symbolic links in bin & lib:
ln -s lx24-amd64 lx26-amd64
This way the lx26-amd64 directories are there (even if they are linked
back to lx24-amd64) when SGE daemons need to reference files in them
in lib & bin.
> unlike what i said previously:
> spooling_method berkeleydb
> spooling_lib libspoolb
So you are using BDB spooling and there's nothing wrong with
BerkeleyDB inherently (the NFSv4 dependency was removed in GE2011.11
so the only thing that classic spooling is better at is mainly the
human editable on-disk files).
> Can i switch to classic spool without losing the accounting ?
Yes, you can switch to classic spooling, and the accounting file is
separate from the spooling framework.
However, you don't need to switch if Berkeley DB has been working fine
for you. Also, keep in mind that switching spooling frameworks is not
And finally, let me reiterate that if all you need is just need the
4GB memory accounting fix, create sym. links for lx26-amd64, then copy
the sge_execd compiled from the SGE6.2u5p2 source, and restart
sge_execd on the nodes.
> thanks for reply,
>> On Wed, Jan 25, 2012 at 6:33 AM, Gerard Henry<ghenry at cmi.univ-mrs.fr>
>>> hello all,
>>> we have the old SGE 6.2u5
>>> I'm trying to compile SGE6.2u5p2, i just need the fix for the 4Go mem
>>> My idea was to recompile binaries, and to replace the actual sge_execd
>>> the new ones.
>>> I follow the instructions here:
>>> everything seems ok. I get:
>>> and i copy these files into my existing SGE installation.
>>> before that, the dirs were:
>>> when re-starting qmaster, it uses lx26 instead of lx24 (i'm on centos )
>>> it fails, and in messages:
>>> 01/25/2012 11:50:17| main|octopus|E|Binaries were compiled with spooling
>>> method "classic", requested spooling method is "berkeleydb"
>>> 01/25/2012 11:50:17| main|octopus|E|error creating a "berkeleydb"
>>> 01/25/2012 11:50:17| main|octopus|C|setup failed
>>> why does it search for berkeleydb ? since i'm using classic spool.
>>> thanks in advance for help,
>>> users mailing list
>>> users at gridengine.org
More information about the users