If you do not have enough
nodes to test, there will be no logs. There is nothing "missing" since the
job never started. You cannot have log files for something that doesn't
exist.
I believe it should be safe
to update the repository, as yume should be smart enough to pick a newer version
when necessary and also update older versions. However, packages are not
the only things that are changed between revisions, some times packages are
updated, but there are corresponding code/scripts that were modified to support
the new RPMs, so simply updating the package will only get you part way
there.
There is currently no
easy way to update these scripts. This will change in
the future. However - you could probably just overwrite what you
have with the ones from the SVN repository, but there are no gurantees whether
that will make unexpected changes to your system.
I know of at least one user
(mjhsieh?) who is using the development code and would like to "update" it to
the production code when it is ready. There will be no official support
for this since the code is development and we expect users to re-install using
the release code when it is ready. Besides, I believe that you are
familiar enough such that it won't take you that long to re-install the whole
thing once more stable code is available.
So my advice to you is "If
it's not broken, don't fix it". You had issues with openmpi and I gave you
a perfectly working solution (which is to make sure that you had the same
version between openmpi and openmpi-switcher-modulefile) that should get you
going further.
Perhaps other developers have
more information they can provide you.
Regards,
Bernard
From: Brad Aisa [mailto:[EMAIL PROTECTED]
Sent: Sun 23/07/2006 17:37
To: oscar devel
Cc: Bernard Li
Subject: Re: errors during cluster test
Ok,
openmpi-switcher-modulefile-1.1-1 is there on the nodes (it was
there.)
The MPI test runs for quite awhile (a minute or two maybe?), with some countdown (like from 59 or something), and when it reaches 0, it says it FAILS. The subsequent steps fail, but I guess it is because of the open job still in pbs.
Fixing these missing test logs sure seems like it would be helpful...
Also, what is the word on whether I should update my installation with the latest Oscar packages?
Pretty anxious to start using my cluster here!!!!
Brad Aisa
baisa at brad-aisa dot com
The MPI test runs for quite awhile (a minute or two maybe?), with some countdown (like from 59 or something), and when it reaches 0, it says it FAILS. The subsequent steps fail, but I guess it is because of the open job still in pbs.
Fixing these missing test logs sure seems like it would be helpful...
Also, what is the word on whether I should update my installation with the latest Oscar packages?
Pretty anxious to start using my cluster here!!!!
baisa at brad-aisa dot com
-----
Original Message ----
From: Bernard Li <[EMAIL PROTECTED]>
To: Brad Aisa <[EMAIL PROTECTED]>; oscar devel <[email protected]>
Sent: Sunday, July 23, 2006 6:22:45 PM
Subject: RE: errors during cluster test
From: Bernard Li <[EMAIL PROTECTED]>
To: Brad Aisa <[EMAIL PROTECTED]>; oscar devel <[email protected]>
Sent: Sunday, July 23, 2006 6:22:45 PM
Subject: RE: errors during cluster test
That's a bug with
TORQUE. However manually deleting the job works.
BTW, I think there's also a
--force option with qdel in newer versions of TORQUE, not sure if it's available
in the version we have (or maybe I got confused with SGE).
Regarding the RPM, it should
be "openmpi-switcher-modulefile", not "openmpi-modulefile". Sorry for the
confusion.
------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys -- and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________ Oscar-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/oscar-devel
