Thanks Nate.  I'll play with that.  Could it be that Galaxy is trying to reset the permissions or ownership of the imported BAM files.  I'm not copying them into Galaxy, rather I am linking to them.  That is the only error I see in runner0.log that indicates any type of failure.

On Fri, Jan 20, 2012 at 10:35 AM, Nate Coraor <nate@bx.psu.edu> wrote:
On Jan 18, 2012, at 11:54 AM, Ryan Golhar wrote:

> Nate - Is there a specific place in the Galaxy code that forks the samtools index on bam files on the cluster or the head node?  I really need to track this down.

Hey Ryan,

Sorry it's taken so long, I've been pretty busy.  The relevant code is in galaxy-dist/lib/galaxy/datatypes/binary.py, in the Bam class.  When Galaxy runs a tool, it creates a Job, which is placed inside a JobWrapper in lib/galaxy/jobs/__init__.py.  After the job execution is complete, the JobWrapper.finish() method is called, which contains:

                   if not self.app.config.set_metadata_externally or \
                    ( not self.external_output_metadata.external_metadata_set_successfully( dataset, self.sa_session ) \
                      and self.app.config.retry_metadata_internally ):
                       dataset.set_meta( overwrite = False )

Somehow, this conditional is being entered.  Since set_metadata_externally is set to True, presumably the problem is external_metadata_set_successfully() is returning False and retry_metadata_internally is set to True.  If you leave behind the relevant job files (cleanup_job = never) and have a look at the PBS and metadata outputs you may be able to see what's happening.  Also, you'll want to set retry_metadata_internally = False.

--nate

>
> On Fri, Jan 13, 2012 at 12:54 PM, Ryan Golhar <ngsbioinformatics@gmail.com> wrote:
> I re-uploaded 3 BAM files using the "Upload system file paths.  runner0.log shows:
>
> galaxy.jobs DEBUG 2012-01-13 12:50:08,442 dispatching job 76 to pbs runner
> galaxy.jobs INFO 2012-01-13 12:50:08,555 job 76 dispatched
> galaxy.jobs.runners.pbs DEBUG 2012-01-13 12:50:08,697 (76) submitting file /home/galaxy/galaxy-dist-9/database/pbs/76.sh
> galaxy.jobs.runners.pbs DEBUG 2012-01-13 12:50:08,697 (76) command is: python /home/galaxy/galaxy-dist-9/tools/data_source/upload.py /home/galaxy/galaxy-dist-9 /home/galaxy/galaxy-dist-9/datatypes_conf.xml /home/galaxy/galaxy-dist-9/database/tmp/tmpqrVYY7         208:/home/galaxy/galaxy-dist-9/database/job_working_directory/76/dataset_208_files:None         209:/home/galaxy/galaxy-dist-9/database/job_working_directory/76/dataset_209_files:None         210:/home/galaxy/galaxy-dist-9/database/job_working_directory/76/dataset_210_files:None; cd /home/galaxy/galaxy-dist-9; /home/galaxy/galaxy-dist-9/set_metadata.sh ./database/files ./database/tmp . datatypes_conf.xml ./database/job_working_directory/76/galaxy.json
> galaxy.jobs.runners.pbs DEBUG 2012-01-13 12:50:08,699 (76) queued in default queue as 114.localhost.localdomain
> galaxy.jobs.runners.pbs DEBUG 2012-01-13 12:50:09,037 (76/114.localhost.localdomain) PBS job state changed from N to R
> galaxy.jobs.runners.pbs DEBUG 2012-01-13 12:51:09,205 (76/114.localhost.localdomain) PBS job state changed from R to E
> galaxy.jobs.runners.pbs DEBUG 2012-01-13 12:51:10,206 (76/114.localhost.localdomain) PBS job state changed from E to C
> galaxy.jobs.runners.pbs DEBUG 2012-01-13 12:51:10,206 (76/114.localhost.localdomain) PBS job has completed successfully
>
> 76.sh shows:
> [galaxy@bic pbs]$ more 76.sh
> #!/bin/sh
> GALAXY_LIB="/home/galaxy/galaxy-dist-9/lib"
> if [ "$GALAXY_LIB" != "None" ]; then
>     if [ -n "$PYTHONPATH" ]; then
>         export PYTHONPATH="$GALAXY_LIB:$PYTHONPATH"
>     else
>         export PYTHONPATH="$GALAXY_LIB"
>     fi
> fi
> cd /home/galaxy/galaxy-dist-9/database/job_working_directory/76
> python /home/galaxy/galaxy-dist-9/tools/data_source/upload.py /home/galaxy/galaxy-dist-9 /home/galaxy/galaxy-dist-9/datatypes_conf.xml /home/galaxy
> /galaxy-dist-9/database/tmp/tmpqrVYY7         208:/home/galaxy/galaxy-dist-9/database/job_working_directory/76/dataset_208_files:None         209:/
> home/galaxy/galaxy-dist-9/database/job_working_directory/76/dataset_209_files:None         210:/home/galaxy/galaxy-dist-9/database/job_working_dire
> ctory/76/dataset_210_files:None; cd /home/galaxy/galaxy-dist-9; /home/galaxy/galaxy-dist-9/set_metadata.sh ./database/files ./database/tmp . dataty
> pes_conf.xml ./database/job_working_directory/76/galaxy.json
>
> Right as the job ended I check the job output files:
>
> [galaxy@bic pbs]$ ll
> total 4
> -rw-rw-r-- 1 galaxy galaxy 950 Jan 13 12:50 76.sh
> [galaxy@bic pbs]$ ll
> total 4
> -rw------- 1 galaxy galaxy   0 Jan 13 12:50 76.e
> -rw------- 1 galaxy galaxy   0 Jan 13 12:50 76.o
> -rw-rw-r-- 1 galaxy galaxy 950 Jan 13 12:50 76.sh
>
> samtools is now running on the head node.
>
>
> Where does Galaxy decide how to run samtools?  Maybe I can add a check of some sort to see what's going on?
>
>
> On Fri, Jan 13, 2012 at 10:53 AM, Nate Coraor <nate@bx.psu.edu> wrote:
> On Jan 12, 2012, at 11:41 PM, Ryan Golhar wrote:
>
> > Any ideas as to how to fix this?  We are interested in using Galaxy to host all our NGS data.  If indexing on the head node is going to happen, then this is going to be an extremely slow process.
>
> Could you post the contents of /home/galaxy/galaxy-dist-9/database/pbs/62.sh ?
>
> Although I have to admit this is really baffling.  The presence of this line without an error:
>
>    galaxy.datatypes.metadata DEBUG 2012-01-11 10:22:40,162 Cleaning up external metadata files
>
> Indicates that metadata was set externally and the relevant metadata files were present on disk.
>
> --nate
>
>
>