So from what I am reading it is either not creating or prematurely deleting this working directory? I will mod that option and see if I can prevent it from deleting.

AC

On 11/12/2012 2:11 PM, Nate Coraor wrote:
On Nov 12, 2012, at 11:17 AM, Alan V. Cowles wrote:

yeah I'm not seeing the folder 11, I have a folder 2 created july 19...


(galaxy_env)-bash-4.1$ pwd
/usr/local/galaxy/galaxy-dist/database/job_working_directory/000


(galaxy_env)-bash-4.1$ ls -al
total 12
drwxrwxr-x 3 galaxy galaxy 4096 Nov  6 14:41 .
drwxrwxr-x 3 galaxy galaxy 4096 Jul 19 11:51 ..
drwxrwxr-x 2 galaxy galaxy 4096 Jul 19 12:47 2


should I make 11? or is it supposed to do this automatically for the job?
A working directory will be created every time a new job runs.  You can use the 
"cleanup_job" config option to control whether this directory is removed after 
job completion.

--nate



On 11/12/2012 10:54 AM, Nate Coraor wrote:
On Nov 6, 2012, at 2:44 PM, Alan V. Cowles wrote:

same error without the symlink, this time with actual syntax:

galaxy.jobs ERROR 2012-11-06 14:41:06,596 (11) Failed to change ownership of 
/usr/local/galaxy/galaxy-dist/database/job_working_directory/000/11, failing
Hi Alan,

You're using the "run jobs as the actual user" code?  It looks like the job's 
working directory (/usr/local/galaxy/galaxy-dist/database/job_working_directory/000/11) 
does not exist after the job finishes running:

Traceback (most recent call last):
   File "/usr/local/galaxy/galaxy-dist/lib/galaxy/jobs/__init__.py", line 293, 
in finish
     self.reclaim_ownership()
   File "/usr/local/galaxy/galaxy-dist/lib/galaxy/jobs/__init__.py", line 693, 
in reclaim_ownership
     self._change_ownership( self.galaxy_system_pwent[0], str( 
self.galaxy_system_pwent[3] ) )
   File "/usr/local/galaxy/galaxy-dist/lib/galaxy/jobs/__init__.py", line 679, 
in _change_ownership
     assert p.returncode == 0
AssertionError
galaxy.objectstore CRITICAL 2012-11-06 14:41:08,863 
/usr/local/galaxy/galaxy-dist/database/job_working_directory/000/11 delete 
error [Errno 2] No such file or directory: 
'/usr/local/galaxy/galaxy-dist/database/job_working_directory/000/11'
Which is why you're getting the errors about changing permissions.

--nate

152.3.172.61 - - [06/Nov/2012:14:41:07 -0400] "POST /library_common/library_item_updates HTTP/1.1" 200 - 
"http://galaxy.igsp.duke.edu:8080/library_common/browse_library?status=done&show_deleted=False&cntrller=library_admin&message=Added+7+datasets+to+the+library+__sq__Omics__sq__+%28each+is+selected%29.++Click+the+Go+button+at+the+bottom+of+this+page+to+edit+the+permissions+on+these+datasets+if+necessary.&id=5969b1f7201f12ae&created_ldda_ids=1462%2C1463%2C1464%2C1465%2C1466%2C1467%2C1468";
 "Mozilla/5.0 (X11; Linux x86_64; rv:16.0) Gecko/20100101 Firefox/16.0"


On 11/06/2012 02:20 PM, Alan V. Cowles wrote:
I tossed the symlink and mounted the san:/vol/galaxy_sata/data/galaxy to 
/usr/local/galaxy, though we use symlinks daily for practically all of our 
softwares and programs practically invisibly, but I am happy to try anything to 
get it going.

AC

On 11/06/2012 02:12 PM, Dave Clements wrote:
Hi Alan,

I'm CC'ing the Galaxy-Dev list on my response to your question.

I talked with Jen Jackson on this.  Her suggestion is that the symlink is most 
likely the problem somehow - this time.  When in doubt, getting rid of these 
good place to start. So recommended starting solution to try - use full paths 
for everything.  Also probably keep all on same partition for the best 
performance, but that would be hard to defend (there was some debate about this 
recently).

Many thanks to Jen for her (as always) careful consideration.

Dave C.


On Mon, Nov 5, 2012 at 7:15 AM, Alan V. Cowles <alan.cow...@duke.edu> wrote:
Hey Dave,

Thanks for getting back in touch with me.

We are running our galaxy instance on a vm... with Centos (RHEL) 6.2, 2 vcpus, 
and 2gb vram, with a remote postgresql prod server as our database instance.

Our first issue was trying to get it so we could load sequence and larger files 
directly from an nfs mounted directory on the server, without uploading via the 
web (php vs large files problem.) We accomplished this via the data libraries 
interface but it's still not very user intuitive.

Our first demo user attempted to upload 3 large files and actually filled up 
the / partition on the server as galaxy was installed at /usr/local.

After scrapping and rebuilding that, we now have a several hundred GB nfs 
partition, mounted and /usr/local/galaxy is a symlink to a folder on this 
volume.

The issue we are currently seeing is during an upload, or even just linking 
files it runs for a while before erroring out and saying there is no 
information on the error. The file seems to be there and taking up the correct 
amount of disk space but galaxy has a red bar.

The galaxy server instance is running back grounded in a screen so we can 
always switch over and look at the output it shows to standard out as errors 
happen, and it seems that it has trouble changing permissions on the files in 
the nfs directory, even though, root squash is disabled, and the entire 
directory is 777 at the top level and everything underneath is owned 
galaxy:galaxy.

AC





--
http://galaxyproject.org/
http://getgalaxy.org/
http://usegalaxy.org/
http://wiki.galaxyproject.org/

___________________________________________________________
Please keep all replies on the list by using "reply all"
in your mail client.  To manage your subscriptions to this
and other Galaxy lists, please use the interface at:

  http://lists.bx.psu.edu/

___________________________________________________________
Please keep all replies on the list by using "reply all"
in your mail client.  To manage your subscriptions to this
and other Galaxy lists, please use the interface at:

 http://lists.bx.psu.edu/

Reply via email to