commit:6caf9c47

This branch fixes the assignment of qsequence values to job tasks. It may not completely fix bug #3113 (that needs more testing) but it's definitely a worthwhile bugfix in its own right.

Bring the docker scripts up to date with * Docker 1.0 * Current versions of Arvados components With a fresh docker install, it should be possible to: * Use a browser (running on a real host) to log in to Workbench * Use arv-put and arv-get (running on a real host) to store/retrieve data in the dockerized Keep services

Notes: * When testing documentation, start by running "putty -cleanup" to remove cached host keys, saved sessions, etc. http://the.earth.li/~sgtatham/putty/0.53b/htmldoc/Chapter3.html#3.7.2 * Include a "troubleshooting" point, testing plink-to-switchyard and at least covering this particular issue: "If black window appears and nothing further happens, open command prompt and type XYZ. It should say XYZ."

Use 2-week permission signature TTL in both Keep and API server.

Changes at commit:975b191: Callbacks to force permission link ownership to system_user_uuid. The before_save callbacks ArvadosModel.ensure_owner_uuid_is_permitted and Link.permission_to_attach_to_objects now test links to see if the current user has permission to manage the head_uuid of the link.

Update integration tests to verify the Help down down. - pipeline instannces or projects tests - anonymous user tests

- Do we want another drop down to the right of gear menu?

After the bug in issue #3031, workbench tries to run the error page, which blows up like this: Rendered application/error.html.erb within layouts/application (1.2ms) Completed 500 Internal Server Error in 664ms ActionView::Template::Error (ArvadosApiClient::NotLoggedInException): 173: <% end %> 174: My projects 175: </li> 176: <% my_project_tree.each do |pnode| %> 177: <% next if pnode[:object].class != Group %> 178: <li style="padding-left: <%= pnode[:depth]-1 %>em"> 179: <%= link_to(pnode[:object].name, project_path(pnode[:object].uuid), data: {object_uuid: pnode[:object].uuid, name: 'name'}) %> app/models/arvados_api_client.rb:88:in `api' app/models/arvados_resource_list.rb:69:in `where' app/models/arvados_resource_list.rb:75:in `results' app/models/arvados_resource_list.rb:88:in `each' app/controllers/application_controller.rb:613:in `build_project_trees' app/controllers/application_controller.rb:600:in `my_project_tree' app/views/layouts/application.html.erb:176:in `_app_views_layouts_application_html_erb___4156834235341399204_21093860' app/controllers/application_controller.rb:46:in `block (2 levels) in render_error' app/controllers/application_controller.rb:40:in `render_error' app/controllers/application_controller.rb:58:in `render_exception'

Are tags saved in the .tar somehow? Or is there some other clear way to save that metadata? We could just put it in the same Keep collection in a separate file…

At its core, this does @docker save [arguments] | arv-put@. It also inspects the metadata in the tar archive, and makes links resources from them to capture that information in Arvados.

This implements the tool described in #2998.

* Add a new @docker_image_locator@ column to the Jobs table. * crunch-job installs images from Keep rather than using docker pull. * crunch-job records the Keep Collection locator used in the new column. * Support using a Keep Collection locator as the @docker_image@ runtime constraint.

This implements the changes discussed in #2923.

This branch implements the work described in #2924. It implements filters-based Job mapping with an eye toward reusing this functionality in other controllers in the future.

Compare the constraint Docker image against the @docker_image_locator@ of the past Job. If they don't match, the Job is not eligible to reuse.