Actually, i did not want to use file staging.  I am running from my home directory which is visible on all nodes, and has plenty of space for my needs.  I'm guessing I did something in the config to activate file staging, in which case that would explain the error that shouldn't be.  I didn't read closely enough, but I set:

# The PBS options are described in detail in the Galaxy Configuration section of
# the ClusteringGalaxy Wiki, and are only necessary when using file staging.
pbs_application_server = variome
pbs_stage_path = /home/perin/galaxy-dist/database/tmp
pbs_dataset_server = variome

which obviously caused issues as a result.  Commenting this back cleared it up, but introduced another issue, where the nodes apparently don't see the binary it is trying to run.  I was assuming that the binaries were located in the home directory I was running from, but now see that they are actually in /usr/local/bin/etc...   I'm not sure where this is set, but assume it was done sometime during the setup.sh process.  Is there a way I can specify this install location, or change it anywhere?  I can copy the binaries over to the nodes, but would prefer to host them in a shared place.  Is there a way to change Galaxy so that it looks in say "/share/apps/bin/" instead of /usr/local/bin ?

Thanks!

Juan


On Fri, Oct 30, 2009 at 2:30 PM, Nate Coraor <nate@bx.psu.edu> wrote:
juan perin wrote:

Thanks for the answer!  I've copied this response to the dev list email instead, after realizing its probably best there.  

Agreed.


I'm also having some trouble getting our PBS system configured.  I have pbs://hostname/ as my pbs parameter in the universe file, as well as everything else specified in the documentation.  I'm getting no error in the interface, and the backend never actually gets the job.  The error I see in the logs says:

galaxy.jobs.runners.pbs ERROR 2009-10-30 13:40:27,762 Uncaught exception queueing job
Traceback (most recent call last):
 File "/home/perin/galaxy-dist/lib/galaxy/jobs/runners/pbs.py", line 144, in run_next
   self.queue_job( obj )
 File "/home/perin/galaxy-dist/lib/galaxy/jobs/runners/pbs.py", line 219, in queue_job
   script = pbs_symlink_template % (job_wrapper.galaxy_lib_dir, " ".join(job_wrapper.get_input_fnames() + output_fnames), self.app.config.pbs_stage_path, exec_dir, command_line)
TypeError: sequence item 1: expected string, DatasetPath found

Hi Juan,

It looks like you're using file staging, and there was a bug in our file staging code (we don't use staging ourselves anymore, so it wasn't caught before).

Here's the fix, you can pull it from our central repo or simply apply it manually:

http://bitbucket.org/galaxy/galaxy-central/changeset/dac542cca894/

--nate