[galaxy-dev] galaxy on ubuntu 14.04: hangs on metadata cleanup
Dear all, Has anyone tried running Galaxy on Ubuntu 14.04? I’m trying a test setup on two virtual machines (worker+master) with a SLURM queue. Getting in strange problems when jobs finish, the master hangs, completely unresponsive with CPU at 100% (as reported by virt-manager, not by top). Only drmaa jobs seem to be affected. After hanging, a reboot shows the job is finished (and green in history). It took me some debugging to figure out where things go wrong, but it seems it goes wrong when os.remove is called in lib/galaxy/datatypes/metadata.py in method cleanup_external_metadata. I can reproduce the problem by calling os.remove(metadatafile) by hand (in an interactive python shell) when using pdb to create a breakpoint just before the call. If I comment out the os.remove it runs on until it hits another delete call in lib/galaxy/jobs/__init__.py: self.app.object_store.delete(self.get_job(), base_dir='job_work', entire_dir=True, dir_only=True, extra_dir=str(self.job_id)) It’s in the JobWrapper class in the cleanup() method. I should mention here that my galaxy version is a bit old since I’m running my own fork with local modifications on datatypes. This object_store.delete also leads to a shutil.rmtree and os.remove function. So, remove calls to the filesystem seem to hang the whole thing, but only at this point in time. Rebooting and removing by hand is no problem, pdb-stepping also sometimes fixes it (but if I just press continue it hangs). I don’t know where to go from here with debugging, but has anyone seen anything similar? Right now it feels like it may be caused by timing rather than actual code problems. cheers, — Jorrit Boekel Proteomics systems developer BILS / Lehtiö lab Scilifelab Stockholm, Sweden ___ 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/ To search Galaxy mailing lists use the unified search at: http://galaxyproject.org/search/mailinglists/
[galaxy-dev] Installation failure on Test Tool Shed
Hi Dave, Can you tell me any more about this failed tool dependency error: http://testtoolshed.g2.bx.psu.edu/view/peterjc/mira4_assembler Installation errors Tool dependencies TypeNameVersion MIRA package 4.0 Error File /var/opt/buildslaves/buildslave-ec2-2/buildbot-install-test-test-tool-shed-py27/build/lib/tool_shed/galaxy_install/tool_dependencies/install_util.py, line 345, in install_and_build_package_via_fabric tool_dependency = fabric_util.install_and_build_package( app, tool_shed_repository, tool_dependency, actions_dict ) File /var/opt/buildslaves/buildslave-ec2-2/buildbot-install-test-test-tool-shed-py27/build/lib/tool_shed/galaxy_install/tool_dependencies/fabric_util.py, line 93, in install_and_build_package initial_download=False ) File /var/opt/buildslaves/buildslave-ec2-2/buildbot-install-test-test-tool-shed-py27/build/lib/tool_shed/galaxy_install/tool_dependencies/recipe/recipe_manager.py, line 418, in execute_step initial_download=initial_download ) File /var/opt/buildslaves/buildslave-ec2-2/buildbot-install-test-test-tool-shed-py27/build/lib/tool_shed/galaxy_install/tool_dependencies/recipe/step_handler.py, line 1312, in execute_step return_output=False ) File /var/opt/buildslaves/buildslave-ec2-2/buildbot-install-test-test-tool-shed-py27/build/lib/tool_shed/galaxy_install/tool_dependencies/recipe/recipe_manager.py, line 247, in handle_command output = self.handle_complex_command( command ) File /var/opt/buildslaves/buildslave-ec2-2/buildbot-install-test-test-tool-shed-py27/build/lib/tool_shed/galaxy_install/tool_dependencies/recipe/recipe_manager.py, line 287, in handle_complex_command cwd=state.env[ 'lcwd' ] ) File /usr/lib/python2.7/subprocess.py, line 711, in __init__ errread, errwrite) File /usr/lib/python2.7/subprocess.py, line 1308, in _execute_child raise child_exception [Errno 2] No such file or directory: '/var/opt/buildslaves/buildslave-ec2-2/buildbot-install-test-test-tool-shed-py27/build/test/install_and_test_tool_shed_repositories/repositories_with_tools/tmp/tmpxFCguW/tmp-toolshed-mtdt0b857/MIRA' I presume this was a Python stack trace but the whitespace has been lost when rendered as HTML. However, I don't quite see what is going wrong here... Thanks, Peter ___ 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/ To search Galaxy mailing lists use the unified search at: http://galaxyproject.org/search/mailinglists/
Re: [galaxy-dev] galaxy on ubuntu 14.04: hangs on metadata cleanup
I should probably mention that the data filesystem is NFS, exported by the master from /mnt/galaxy/data and mounted on the worker. No separate fileserver. Master is the one that hangs. cheers, — Jorrit Boekel Proteomics systems developer BILS / Lehtiö lab Scilifelab Stockholm, Sweden On 07 May 2014, at 15:57, Jorrit Boekel jorrit.boe...@scilifelab.se wrote: Dear all, Has anyone tried running Galaxy on Ubuntu 14.04? I’m trying a test setup on two virtual machines (worker+master) with a SLURM queue. Getting in strange problems when jobs finish, the master hangs, completely unresponsive with CPU at 100% (as reported by virt-manager, not by top). Only drmaa jobs seem to be affected. After hanging, a reboot shows the job is finished (and green in history). It took me some debugging to figure out where things go wrong, but it seems it goes wrong when os.remove is called in lib/galaxy/datatypes/metadata.py in method cleanup_external_metadata. I can reproduce the problem by calling os.remove(metadatafile) by hand (in an interactive python shell) when using pdb to create a breakpoint just before the call. If I comment out the os.remove it runs on until it hits another delete call in lib/galaxy/jobs/__init__.py: self.app.object_store.delete(self.get_job(), base_dir='job_work', entire_dir=True, dir_only=True, extra_dir=str(self.job_id)) It’s in the JobWrapper class in the cleanup() method. I should mention here that my galaxy version is a bit old since I’m running my own fork with local modifications on datatypes. This object_store.delete also leads to a shutil.rmtree and os.remove function. So, remove calls to the filesystem seem to hang the whole thing, but only at this point in time. Rebooting and removing by hand is no problem, pdb-stepping also sometimes fixes it (but if I just press continue it hangs). I don’t know where to go from here with debugging, but has anyone seen anything similar? Right now it feels like it may be caused by timing rather than actual code problems. cheers, — Jorrit Boekel Proteomics systems developer BILS / Lehtiö lab Scilifelab Stockholm, Sweden ___ 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/ To search Galaxy mailing lists use the unified search at: http://galaxyproject.org/search/mailinglists/
[galaxy-dev] Correct toolshed tool config for a .loc file
I'll try to keep this short! I'm wondering why on a fresh install of galaxy via a setup script into a python virtualenv, my custom toolshed tool's .loc files don't seem to be loading, i.e. the tool's tool_data_table_conf.xml file isn't loading. All the other parts of the tool (and other toolshed tools) do run ok. On the galaxy install where the tool was developed, it runs ok too, and I can't see any universe.wsgi.conf differences that would account for this though I'm sure the dev server has a number of other different config settings. I thought I knew my way around this .loc stuff, but I'm flummoxed. It is as though /usr/local/galaxy/test/galaxy-dist/tool-data/[toolshed domain etc etc]/00d397a381f8/tool_data_table_conf.xml isn't loading into Galaxy's tabular data list?! The only other thing I can think of is that I stuffed the tool-data related loc files into a subfolder of the tool blast_reporting.py blast_reporting.xml tool_data_table_conf.xml.sample ... tool-data/ blast_reporting_fields.loc.sample But the resulting downloaded tool config in the galaxy-dist/tool-data/. /blast_reporting_fields.loc and tool_data_table_conf.xml files seem just fine!!! Regards, Damion ___ 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/ To search Galaxy mailing lists use the unified search at: http://galaxyproject.org/search/mailinglists/
[galaxy-dev] setup_virtualenv action doesn't update PATH
Hello List, I'm writing a tool_dependencies.xml file for a python package. It defines an action type=setup_virtualenv, and when I install it to my galaxy from the test shed it is actually installed into ~/environments, and the tool definition is set up correctly. This package includes a callable script, when I install the package on a virtualenv using pip, the path to the CLI tool is inserted to the PATH and I can call it anywhere. But when I run the tool from galaxy it cannot find the script. Does the setup_virtualenv action also update the PATH? Thanks! Rodrigo ___ 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/ To search Galaxy mailing lists use the unified search at: http://galaxyproject.org/search/mailinglists/
Re: [galaxy-dev] Correct toolshed tool config for a .loc file
One further piece of the mystery I see now: The View Data tables Registry shows my tool's loc files, but has odd stuff for Tool data path and Missing index file columns, for example: blast_reporting_fields [tab] /usr/local/galaxy/test/galaxy-dist/tool-data/salk.bccdc.med.ubc.ca/toolshed/repos/ddooley/blast_reporting/00d397a381f8/blast_reporting_fields.loc [tab] /usr/local/galaxy/test/galaxy-dist/tool-data ./database/tmp/tmp-toolshed-gmfcrDITxyK/blast_reporting_fields.loc [tab] ./database/tmp/tmp-toolshed-gmfcrDITxyK [tab] missing There is no galaxy-dist/database/tmp-toolshed-gmfcrDITxyK folder, if that's where its supposed to be? Is there some setting for database tmp folder cache for tools? Thanks for tips ... Damion ___ 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/ To search Galaxy mailing lists use the unified search at: http://galaxyproject.org/search/mailinglists/
[galaxy-dev] FW: Correct toolshed tool config for a .loc file - addendum
So I see this now in main.log: galaxy.tools.data WARNING 2014-05-07 16:40:29,613 Line 3 in tool data table 'blast_reporting_fields' is invalid (HINT: 'TAB' characters must be used to separate fields): length numeric int 1 1 1 Alignment length The thing is, tab characters ARE separating the fields. Its just that some columns don't have values. Now, this was working fine on the other galaxy install, updated to December 2013 changeset 11247 . So has .loc processing changed since then? Damion ___ 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/ To search Galaxy mailing lists use the unified search at: http://galaxyproject.org/search/mailinglists/
[galaxy-dev] FW: Correct toolshed tool config for a .loc file - RESOLVED
Well Galaxians, I have to eat humble pie, kind of. Mismatches between the tool's tool_data_table_conf.xml and the actual tabular data were responsible for the disappearing act, which log rather indirectly alluded to. Something that didn't show itself on the dev server. I'll push for some greater reporting on this though, on the Trello board. What I'd like to see is the View data tables registry report have its Missing index file column changed to something like Status and under that you get to see Missing index file or Bad file - Parse error, 23 lines etc. So my apologies to anyone who put some time into this. Damion ___ 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/ To search Galaxy mailing lists use the unified search at: http://galaxyproject.org/search/mailinglists/