Hello Björn, On Apr 6, 2013, at 9:06 AM, Björn Grüning wrote:
Hi Greg and Dave,
good to know that it is at least installing :) I also tried without apache, fresh checkout and get the same timeouts. Maybe my computer is just so slow that its triggers the timeout. Also my Memory is very limited and nearly swapped during compilation.
But with the new API I'm able to install all of them just fine. Is that explainable? During my experiments I found that im not able to install the tools if a new-panel-section is supplied:
galaxy.web.framework ERROR 2013-04-06 14:29:05,411 Uncaught exception in exposed API method:
Traceback (most recent call last): File "/home/bag/projects/galaxy/galaxy-central/lib/galaxy/web/framework/__init__.py", line 196, in decorator rval = func( self, trans, *args, **kwargs) File "/home/bag/projects/galaxy/galaxy-central/lib/galaxy/webapps/galaxy/api/tool_shed_repositories.py", line 254, in install_repository_revision reinstalling=False ) File "/home/bag/projects/galaxy/galaxy-central/lib/tool_shed/galaxy_install/repository_util.py", line 358, in install_tool_shed_repository if tool_panel_section_key: File "/usr/lib/python2.7/UserDict.py", line 23, in __getitem__ raise KeyError(key) KeyError: 'section_chemicaltoolbox'
This issue should be fixed in revision 3a07307 - thanks for reporting this!
Without that option it works. I tried to understand the code-path but found no section where a new entry is created under the toolbox. What I found is that: tool_panel_section_keys in repository_dependecy_util.py line 235 is updated every time and you will end up with several identical instances of 'section_chemicaltoolbox'. Hope that helps a little bit.
I'm not able to reproduce this (assuming it is not related to the issue I fixed above). Is this issue now resolved for your?
Also please see a pull-request for a small typo: https://bitbucket.org/galaxy/galaxy-central/pull-request/150/change-error-me...
This pull request has been accepted - thanks for the fixes!
Thanks for your work! Bjoern
Hello Björn,
Dave Bouvier (cc'd on this) spent some time yesterday inspecting the logs for out test tool shed and found no timeouts in either our paster log or our nginx log. Dave was also successful in installing your repository and dependencies. If I remember correctly, this timeout issue may be intermittent, but again, Dave found no logged timeout events at our end. If you uncover any additional information, please let us know, and we'll continue to keep an eye on this.
Thanks very much,
Greg Von Kuster
On Apr 4, 2013, at 5:32 AM, Björn Grüning wrote:
Hi Greg,
sorry forgot to attach the new traceback. Its now attached. If the grepping with the IP address did not work, maybe you can grep for the chemicaltoolbox/chemfp/openbabel/osra repositories. Im probably the only one working with it.
Have a nice day! Bjoern
Hello Björn,
Thanks very much for your message! Please see my inline comments.
On Apr 2, 2013, at 4:33 PM, Björn Grüning wrote:
I have tested a few things regarding to the toolshed and prepared a complex module-dependency-beast, with meta-package and orhphan-tool-dependencies and so on. A few problems I encountered are summarized below. You can try it by yourself, if you install the chemicaltoolbox package from the testtoolshed.
- three crashes with traceback attached, during tool-installation
- one 'name' not defined error, with attached patch
Thanks very much for finding this issue and for the patch. We've applied it in changeset 9252:da2d6ee29f30 in our stable branch, so it is running on both public Galaxy tool sheds.
- For me the the installation of tools with really complex and long compiling steps are stalled in 'Installing' forever. Try to install the osra package (it often hangs, but not every time). Its not possible to uninstall/reinstall these dependencies without touching the database. And the error says, tool is already installed. Would be nice if the admin is able to reset that repository easily, even if it is limbo ... (is there a timeout somewhere?) Error: paste.httpserver.ThreadPool WARNING 2013-04-02 22:02:41,256 Thread 140246687659776 hung (working on task for 1964 seconds) (The SystemExit: traceback attached)
I've create the following Trello card for this issue:
https://trello.com/card/toolshed-enable-galaxy-admin-to-reset-tool-dependenc...
There are 4 places where this timeout may be occurring: Galaxy tool shed server, Galaxy tool shed nginx, a proxy server at your end, or your local Galaxy server. There is not enough information in your attached traceback log to determine where the timeout is occurring. Can you provide me your Galaxy server's IP address so I can look in logs at our end?
- It would be great if the tool-developer or the admin has more control over the location of the tools in the toolbar. Installing meta package leaves you with only one choice, regardless how many tools are bundled.
Support for this will be available in the next Galaxy release. There are Trello cards for this here:
https://trello.com/card/toolshed-enhance-the-current-galaxy-api-for-installi... https://trello.com/card/toolshed-add-features-that-leverage-the-galaxy-api-f...
- Installing dependencies will leave me with empty folders under galaxy-root with the name of the dependency. My guess is, that if I manually download a file/folder with wget, the folder is not purged, but the content is.
Our intent is that installing repositories from the tool shed is completely automated with no manual intervention required. Tool dependency definitions should not use utilities like wget to download dependencies because these utilities may not be available in the environments (e.g. vanilla mac os) into which the repositories are being installed. I'll enhance the tool dependency tag sets as necessary to accommodate additional download and install requirements. Can you let me know if the current download_by_url needs enhancements?
- Often after upgrading the tool_dependencies.xml file, I was not able to 'get updates' from the toolshed. The version was up-to-data, but it wasn't. Deinstalling and Installing fixed it. But I think its ab bug, isn't it?
This very well may be a bug. Can you please provide additional details so that I can reproduce this behavior? I'll get a fix asap.
- Is there a way, to just download a python file, not zipped and add it to PYTHONPATH? If I download it with <action type="download_by_url"> I have problems to copy that file with move. I would like to download it with wget, but one download by URL is required. My current solution is, to include a dummy download_by_url and than fetch with wget, or download a archive version.
I've created the following Trello card for this issue:
https://trello.com/card/toolshed-enhance-the-download-by-url-process-for-ins...
- Is it correct that a repository referenced only in tool_dependencies.xml is not enough to install? numpy was only referenced as tool_dependency and I got the error that the tool (with owner and revision) was not found. After adding to repository_dependency it was ok.
Can you provide more details about your question above (e.g., the tool_dependencies.xml fie would help). I want to make sure i understand the issue.
Thanks very much Björn,
Greg Von Kuster
<trackback.txt>