[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Condor-users] Java Universe and logging query

Any takers for this question?


On 6/20/07, Esh Esh <eshforcondor@xxxxxxxxx> wrote:
Thanks Alan. This information is really useful for me.

With regard to your response for my first query... Does it mean that Vanilla Universe can be used to run any Java job instead of using a Vanilla Universe. In that case what is the necessity of a Java universe when the same can be accomplished with Vanilla Universe. Is there any specific advantage of using a Java universe compared to the Vanilla universe? Are there any tradeoffs?

I have one more related question.

I have multiple machines running in the condor pool. If I need want all these machines to use files / jars located in a central repository for the execution of java job in Java universe, then how can I execute these jobs without having to transfer the required files everytime?


On 6/20/07, Alan De Smet < adesmet@xxxxxxxxxxx > wrote:
Esh Esh < eshforcondor@xxxxxxxxx> wrote:
> 1) Does Condor support job submissions to multiple versions of
> JVM in the Java Universe? Certain jobs might need one version
> of Java and some might need another version of Java. If this is
> not support, any workaround available? Also any plans to
> support it in the future releases?

Condor startds advertise the JVM they are aware of using
JavaVersion in their ClassAd, so jobs can distinguish based on
that.  However, Condor does not directly support multiple Java
installations on a given machine.  You could add your own
attributes ("JAVA_PATH_1_5=/opt/java-1.5/bin/java"
"JAVA_PATH_1_4=/opt/jre-1.4.3/bin/java"), then submit vanilla jobs
that make use of that information (perhaps setting your
executable to a script that relies on getting the path via the
evironment or command line. "arguments=$$(JAVA_PATH_1_4)"

> 2) Is there any different format for logging in condor other
> than text and xml? Further, I am also interested in knowing the
> different events that are logged. Is this configurable in any
> file?

I assume you're referring to the user logs (that is, the log
written when you say "log=myjob.log" in your submit file).  If
that's the case, as you say, it's XML or text.  Both are
reasonably easy to parse.

If you're working in C++ or another language that can work
with C++ libraries, you might look to include/user_log.README in
your Condor install.  It documents our provided API for reading
the log.  include/condor_event.h documents all of the events we
might write.

> 3) We are trying to do a callback from condor for status
> monitoring. Apart from e-mail, is there any other option like
> database, where the status can be checked for any jobs?

Currently there is no callback available.  However, there is a
database option.  Check out Quill:

Alan De Smet                              Condor Project Research
adesmet@xxxxxxxxxxx                 http://www.condorproject.org/
Condor-users mailing list
To unsubscribe, send a message to condor-users-request@xxxxxxx.edu with a
subject: Unsubscribe
You can also unsubscribe by visiting

The archives can be found at:
https://lists.cs.wisc.edu /archive/condor-users/