Hi - I recently discovered Galaxy and am just getting started with it...
I installed a local instance to test with NGS data.
We have a lot of NGS datasets (both Solid and illumina) on a SAN
available via NFS. Each user that has data has an account on the system
in a directory tree such as:
where user1 and user2 are distinct users with their own dataset, and
project1, project2 contain datasets shared by multiple users.
I want to import this data in Galaxy and I came across this thread in
the mailing list (quoted below).
I set up my Galaxy instance to see this NFS share...btw Galaxy is
running as its own user on a virtual machine. The Galaxy user only has
read access to this read and nothing else.
When I went to import the files by specifying a local system path on the
admin user interface, I got errors importing the data. I checked the
paster.log file and I saw errors related to galaxy trying to change the
permissions of the files to 0644.
Does this mean all the files need to be owned by Galaxy?
The best approach to handle this is to have a Galaxy administrator
upload the files into a library, creating library datasets. Set the
following config setting to point to the NFS accessible directory you
use to contain the files:
# Directories of files contained in the following directory can be
uploaded to a library from the Admin view
library_import_dir = /var/opt/galaxy/import
You can set ACCESS permissions on the library to restrict access to
specific users, or leave it public to allow anyone to access the library
Users can import the library datasets into their own histories for
analysis - doing this does not create another disk file, but simply a
"pointer" to the library dataset's disk file.
Greg Von Kuster
Galaxy Development Team
Andreas Kuntzagk wrote:
to reduce traffic when importing big dataset into our local galaxy we
would like to copy directly from the fileserver (which is accessible
from the galaxy server via NFS) into galaxy without moving through the
users desktop computer.
Is there already such a tool by chance? At a glance I could only find
import.py which imports only from a predefined set of files.
So if no such tool exist what would be the best starting point?
For security reasons I would restrict the import to a certain directory.
Anything else to keep in mind?
galaxy-user mailing list
galaxy-user at bx.psu.edu