Hello,

Your're right, there's almost certainly a problem with the
scheduler-event-generator.
I've had some issues with this adaptor too. I'll just copy and paste some
documentation i wrote on that, I'm a little short on time right now so this
is not really pretty, I'm sorry. Tell me if this helped.

Sun Grid Engine Adaptor
If the job appears to execute cleanly in the SGE, but seems to hang at:
"Current job state: Unsubmitted" for the globus toolkit after a while, even
though SGEs qmon or similar lists the job as finished, the Globus Scheduler
Event Generator (SEG), which is part of the adaptor package, is not working
properly.
ps -ef |grep globus-scheduler-event-generator
globus   27828 27773  0 15:08 pts/2    00:00:00
/opt/globus/libexec/globus-scheduler-event-generator -s fork -t 1171021209
globus   27842 27773  0 15:08 pts/2    00:00:00
/opt/globus/libexec/globus-scheduler-event-generator -s sge -t 1170943291
If there is no event generator for "sge", try symlinking
$GLOBUS_LOCATION/lib/*_sge_gcc64dbg.* to $GLOBUS_LOCATION/lib/*_sge_gcc64.*
or similar and restart the globus container. There seems to be a bug where
the debug libraries are not recognized properly.
ln -s libglobus_seg_sge_gcc64dbg.a libglobus_seg_sge_gcc64.a
ln -s libglobus_seg_sge_gcc64dbg.la libglobus_seg_sge_gcc64.la
ln -s libglobus_seg_sge_gcc64dbg.so libglobus_seg_sge_gcc64.so
ln -s libglobus_seg_sge_gcc64dbg.so.0 libglobus_seg_sge_gcc64.so.0
ln -s libglobus_seg_sge_gcc64dbg.so.0.0.0 libglobus_seg_sge_gcc64.so.0.0.0
If the process is running, check it's settings:
cat /opt/globus/etc/globus-sge.conf
log_path=/opt/N1GE6/T-BIGridBC/common/reporting
Also, this logfile should contain a lot of entries with job state changes. if
it only contains a few lines, the sgedbwriter service may clear the logfile
and thus prevents the Scheduler Event Generator from monitoring the jobs. The
dbwriter service has been disabled to avoid this issue.

You can also try to start the event generator manually via
globus-scheduler-event-generator -s sge
And see what it does.

There's also a commercial adaptor from gridwisetech
(http://www.gridwisetech.com/) that also provides a WebMDS Information
provider if you need that one (for gridwise for example). It's also available
under an academic licence.

Regards,
Christian Assfalg


-----Ursprüngliche Nachricht-----
Von: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] Im Auftrag
von Patrick Heckeler
Gesendet: Montag, 20. August 2007 18:05
An: [email protected]
Betreff: [gt-user] Sun Grid Engine GT4 Integration


Hello everybody,

I have succesfully installed the GT-4.0.5 from source and the Sun Grid
Engine 6.1u2 on a ubuntu feisty server (kernel 2.6.20-15-server). Both
work fine.
Then I installed the SGE-Integration packages
(http://www.lesc.ic.ac.uk/projects/SGE-GT4.html).
A native submit (/grid/globus-4.0.5/bin/globusrun-ws -submit  -c
/bin/echo "hello world") is executed with no errors.
But when I submit a job to the Grid Engine
(/grid/globus-4.0.5/bin/globusrun-ws -submit -factory myserver -Ft SGE
-streaming -so /home/username/result -c /bin/echo "hello world" ), I
get following output:
Delegating user credentials...Done.
Submitting job...Done.
Job ID: uuid:f1ee5294-4f33-11dc-9231-000c29eca51d
Termination time: 08/21/2007 15:42 GMT
... an nothing more happens!!!

Maybe there is a problem with the reporting file of the SGE...
I configured the SGE with reporting file (qconf -mconf;
reporting_file=true) and I have activated reporting-variables
(report_variables cpu,np_load_avg,mem_free,virtual_free).
Is there anything else to do in configuration of the SGE?

When I use the "qstat -f "-command, i can see the job beeing executed
by the grid engine correctly. It seems that the communication GT4 ->
SGE works fine... but the other direction  SGE -> GT4 seems not to
work,,,

Does anybody had this problem, too?
Does anybody have an idea how to solve this problem?

Many thanks in advance!

-- 
Viele Grüße / Kind regards

Patrick Heckeler

Teckstr. 6
72074 Tübingen
Germany


Mobil: +49172 / 719 23 24
Homezone: +497071 / 28 47 33
email: pheckeler [at] gmail [dot] com
skype: p_heckeler
ICQ: 200910972

Reply via email to