Hello all,
I've found the solution: The SGE Environment needs to be set for the User
(probably process) that starts the globus container.
Regards,
Christian
-----Ursprüngliche Nachricht-----
Von: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] Im
Auftrag von [EMAIL PROTECTED]
Gesendet: Freitag, 10. August 2007 12:49
An: [email protected]
Betreff: [gt-user] WebMDS not able to process SGE queue information
Hello all,
I've got an error with Globus WebMDS, the Sun Grid Engine scheduler
and the GridWise Tech GE-GT Adaptor, specifically the component that should
provide SGE queue information to WebMDS. This is in the container.log:
2007-08-10 09:31:27,183 WARN
transforms.GLUESchedulerElementTransform [Timer-6,transformElement:377]
Unhandled exception during GLUE ComputeElement transform
ation
java.lang.Exception: Batch provider generated no useful information.
at
org.globus.mds.usefulrp.rpprovider.transforms.GLUESchedulerElementTransform.t
ransformElement(GLUESchedulerElementTransform.java:121)
at
org.globus.mds.usefulrp.rpprovider.TransformElementListener.executionPerforme
d(TransformElementListener.java:81)
at
org.globus.mds.usefulrp.rpprovider.ResourcePropertyProviderTask.timerExpired(
ResourcePropertyProviderTask.java:155)
at
org.globus.wsrf.impl.timer.TimerListenerWrapper.executeTask(TimerListenerWrap
per.java:65)
at
org.globus.wsrf.impl.timer.TimerListenerWrapper.run(TimerListenerWrapper.java
:82)
at java.util.TimerThread.mainLoop(Timer.java:512)
at java.util.TimerThread.run(Timer.java:462)
Aparently, the component that should provide this information does
not work correctly. I've got two GT 4.0.5 Installations that are more or less
similar, so I can compare things. Both stream to one WebMDS, and the Service
Group Entry Details are there for both Hosts. But only one Host has GLUECE
queue information in the AggregatorData section, the Host with this error
does not.
On the host without the error, the WebMDS page section looks like
this:
Content
* AggregatorConfig:
o GetResourcePropertyPollType:
+ PollIntervalMillis: 60000
+ ResourcePropertyName: glue:GLUECE
* AggregatorData:
o GLUECE:
+ ComputingElement:
# Name: all.q
# UniqueID: all.q
# Info:
* GRAMVersion: 4.0.5
* HostName: bibgrid1lx.eu.boehringer.com
* LRMSType: SGE
* LRMSVersion: 6.0u9
* TotalCPUs: 4
[...]
On the host with the error, the AggregatorData section is empty. I
think that this information is provided by a Perl script called
globus-scheduler-provider-sge. This script, however, works on both Hosts, and
produces (almost) identical output like this:
<scheduler xmlns="http://mds.globus.org/batchproviders/2004/09";>
<Info LRMSType="SGE" LRMSVersion="6.0u9" GRAMVersion="4.0.5"
HostName="bibgrid2lx" TotalCPUs="8"/>
<Queue name="all.q">
<totalnodes>4</totalnodes>
<freenodes>4</freenodes>
<maxCount>4</maxCount>
<maxtime>unknown</maxtime>
<maxCPUtime>unknown</maxCPUtime>
<maxReqNodes>unknown</maxReqNodes>
<maxRunningJobs>unknown</maxRunningJobs>
<maxJobsInQueue>unknown</maxJobsInQueue>
<maxTotalMemory>unknown</maxTotalMemory>
<maxSingleMemory>unknown</maxSingleMemory>
<whenActive>unknown</whenActive>
<dispatchType>batch</dispatchType>
<status>enabled</status>
<numJobs>0</numJobs>
</Queue>
<Queue name="restricted.q">
<totalnodes>4</totalnodes>
<freenodes>4</freenodes>
<maxCount>4</maxCount>
<maxtime>unknown</maxtime>
<maxCPUtime>unknown</maxCPUtime>
<maxReqNodes>unknown</maxReqNodes>
<maxRunningJobs>unknown</maxRunningJobs>
<maxJobsInQueue>unknown</maxJobsInQueue>
<maxTotalMemory>unknown</maxTotalMemory>
<maxSingleMemory>unknown</maxSingleMemory>
<whenActive>unknown</whenActive>
<dispatchType>batch</dispatchType>
<status>enabled</status>
<numJobs>0</numJobs>
</Queue>
</scheduler>
My best guess right now is, that I am missing some XML or XSLT
related Library, so that whatever component should parse this output is
unable to do so.
Any suggestions to what might cause this issue, which components are
involved, or where I might get more information on what's going on here?
Mit freundlichen Grüßen,
Christian Aßfalg
BA-Student Informationstechnik
AIV-ITB-SYS2 bei Frau Zocher
