New install and handlers for tools
Galaxy devs, I have just setup a Galaxy instance here in Cardiff University. We have used a VM to host the web frontend and setup DRMAA to connect to our HPC resource. We are still in the process of testing all the functionality it but one aspect which feels a bit strange is how to tell Galaxy to use the web frontend to run the tools which download data. Our HPC resource does not have a connection to the outside world (only the login nodes). I am not sure on the latest version but it seems handlers for tools can only be specified by id not by tool_type or any other "class" descriptor so non-computational tools can be run elsewhere whilst the computational tasks are run on HPC - I am currently using the id for each tool to specify a destination in the job_conf.xml. Am I missing something obvious? Another aspect I found whilst installing this was the delicate nature of permissions when integrating it with the cluster. This is especially true if ACLs are also used on the filesystem. It seems the versions I installed (a version from November) could fail without much information why it failed in the logs. Also had occasions where Galaxy never found out a job failed even though the log file had an error and traceback (usually when the DRMAA submit failed). Any advice (or documentation) what the disk permissions should be? And if anyone else is using NFS then root_squash on NFS really does cause issues. So after a bit of tweaking and getting used to the structure of the code I got something which seemed to work - I have possible a few fixes which others might find useful mainly to give better feedback in the log. Any advice how best to integrate these back to you - is there a ticketing system to track changes? I must really upgrade Galaxy to a more recent version to see whether these have already been fixed. Many thanks for all the work that goes into Galaxy and look forward collaborating with the community. All the best, Tom -- Thomas Green Senior Programmer ARCCA, Redwood Building (Tower), King Edward VII Avenue, Cardiff, Wales, CF10 3NB, UK Tel: +44 (0)29 208 79269 Fax: +44 (0)29 208 70734 Email: greent10@cardiff.ac.uk Web: http://www.cardiff.ac.uk
On Mon, Feb 17, 2014 at 10:26 AM, Thomas Green <GreenT10@cardiff.ac.uk> wrote:
Galaxy devs,
...
So after a bit of tweaking and getting used to the structure of the code I got something which seemed to work - I have possible a few fixes which others might find useful mainly to give better feedback in the log. Any advice how best to integrate these back to you - is there a ticketing system to track changes? I must really upgrade Galaxy to a more recent version to see whether these have already been fixed.
If you're comfortable with hg and bitbucket, submit a pull request: https://bitbucket.org/galaxy/galaxy-central/pull-requests Galaxy used to use the bitbucket issue tracker but switched to Trello for better project management, see: https://wiki.galaxyproject.org/Issues Peter
participants (2)
-
Peter Cock
-
Thomas Green