Brett Smith [Fri, 30 May 2014 14:45:02 +0000 (10:45 -0400)]
Merge branch '2752-arv-put-resume'
Closes #2752, #2864.
Brett Smith [Fri, 30 May 2014 14:21:59 +0000 (10:21 -0400)]
2752: arv-put doesn't update the resume cache when aborted.
It's appealing to try to save the latest state when we have the
opportunity, but the problem is that we can't be sure that our
CollectionWriter is in a consistent state when we get a signal or
exception. The previous code could potentially write inconsistent
state, which would appear to resume successfully but PUT different
data. Instead rely exclusively on cache updates that are done from a
known consistent state.
Brett Smith [Fri, 30 May 2014 13:49:34 +0000 (09:49 -0400)]
2752: arv-put explains resumed uploads in more detail.
Wording suggested by Peter in refs #2752.
Brett Smith [Fri, 30 May 2014 13:46:08 +0000 (09:46 -0400)]
2752: arv-put writes newline after 100% human progress report.
Brett Smith [Fri, 30 May 2014 13:45:24 +0000 (09:45 -0400)]
2752: arv-put always displays 100% progress at the end.
It's okay to call writer.finish_current_stream() this way, because
every branch at the end ultimately calls it.
Brett Smith [Fri, 30 May 2014 13:08:50 +0000 (09:08 -0400)]
2752: Avoid ResumeCache name collisions from concatenating paths.
Brett Smith [Thu, 29 May 2014 20:36:30 +0000 (16:36 -0400)]
2752: arv-put checkpoints after a PUT to Keep.
This allows us to recover after a pulled-the-plug kind of failure, but
is still noticeably less overhead than checkpointing after each file.
Brett Smith [Thu, 29 May 2014 20:09:36 +0000 (16:09 -0400)]
2752: arv-put handles exit signals.
Brett Smith [Thu, 29 May 2014 20:36:14 +0000 (16:36 -0400)]
2752: Resumed collection writer doesn't do_queued_work immediately.
As noted in the comments, callers that build a writer from resumed
state must do_queued_work on it before anything else. But this split
makes it easier to treat initialization problems and work problems
separately, which is critical.
This required refactor progress reporting a bit.
Brett Smith [Thu, 29 May 2014 17:53:29 +0000 (13:53 -0400)]
2752: Remove unused CollectionWriter checkpoint hook.
Brett Smith [Thu, 29 May 2014 17:07:23 +0000 (13:07 -0400)]
2752: arv-put only caches state on failure.
This is slightly less robust, but writing the cache after every data
flush is too expensive.
Brett Smith [Thu, 29 May 2014 17:02:14 +0000 (13:02 -0400)]
2752: Don't duplicate arv-put work after resume.
This change serializes the command-line arguments that we've actually
processed, vs. not. That allows us to safely iterate over them again
to upload any files that we hadn't started, while skipping the ones
we've already done.
Brett Smith [Thu, 29 May 2014 12:55:25 +0000 (08:55 -0400)]
2752: arv-put displays progress quickly after resuming.
Refs #2752.
Brett Smith [Thu, 29 May 2014 12:54:33 +0000 (08:54 -0400)]
2752: arv-put displays a notice when it resumes.
Based on a suggestion from Peter in refs #2752.
Brett Smith [Wed, 28 May 2014 21:27:47 +0000 (17:27 -0400)]
2752: arv-put ResumeCache constructor only accepts path argument.
Refs #2752. Using exceptions to accept parsed arguments in the
constructor as well was too clever by half.
Brett Smith [Wed, 28 May 2014 17:56:10 +0000 (13:56 -0400)]
2752: Fix arv-put's write percentage output.
Brett Smith [Wed, 28 May 2014 17:55:43 +0000 (13:55 -0400)]
2752: Quiet arv-put stdin integration test.
Brett Smith [Tue, 27 May 2014 14:49:13 +0000 (10:49 -0400)]
2752: Avoid hiccup on arv-put with short stdin.
Trying to call os.stat(os.path.realpath('/dev/stdin')) usually fails
in this case. Since we're never going to be able to resume from stdin
anyway, don't bother with that check in this case.
Brett Smith [Tue, 27 May 2014 14:20:50 +0000 (10:20 -0400)]
2752: Improve arv-put initialization from cache.
This puts ArvPutCollectionWriter's own instance variables, like the
progress reporter function, in place before work is resumed.
Brett Smith [Tue, 27 May 2014 14:04:19 +0000 (10:04 -0400)]
2752: arv-put saves and restores write progress information.
This will make output less confusing to the user, and help them
understand that an upload is resuming.
Brett Smith [Mon, 26 May 2014 11:31:07 +0000 (07:31 -0400)]
2752: arv-put shouldn't resume from expired Keep locators.
Brett Smith [Mon, 26 May 2014 11:15:37 +0000 (07:15 -0400)]
2752: Add KeepLocator class to Python SDK.
I hope this can be one place to parse and manipulate locator strings.
Brett Smith [Mon, 26 May 2014 12:12:25 +0000 (08:12 -0400)]
2752: arv-put resumes interrupted downloads from cache.
This commit simply uses all the state-saving work in arv-put's main()
function.
Brett Smith [Fri, 23 May 2014 20:46:49 +0000 (16:46 -0400)]
2752: Add ResumeCache.restart().
This will make it easier for the user to bypass the cache.
Brett Smith [Fri, 23 May 2014 20:22:39 +0000 (16:22 -0400)]
2752: Refactor progress reporting in arv-put's CollectionWriter.
Brett Smith [Fri, 23 May 2014 19:52:50 +0000 (15:52 -0400)]
2752: Separate and test progress reporting functions.
Brett Smith [Fri, 23 May 2014 18:55:56 +0000 (14:55 -0400)]
2752: Factor out arv-put byte tally, and test.
Brett Smith [Fri, 23 May 2014 17:16:25 +0000 (13:16 -0400)]
2752: Add ResumableCollectionWriter serialization to arv-put.
Brett Smith [Mon, 26 May 2014 12:02:15 +0000 (08:02 -0400)]
2752: Add ResumeCache to arv-put.
This class encapsulates all the functionality necessary to
de/serialize ResumableWriter state.
Brett Smith [Mon, 26 May 2014 12:01:27 +0000 (08:01 -0400)]
2752: Generate cache filenames from arv-put arguments.
Brett Smith [Thu, 22 May 2014 21:10:40 +0000 (17:10 -0400)]
2752: Add basic test for arv-put.
Brett Smith [Thu, 22 May 2014 20:46:15 +0000 (16:46 -0400)]
2752: Extract common Python test methods.
I'm going to reuse these to test arv-put.
Brett Smith [Thu, 22 May 2014 20:11:10 +0000 (16:11 -0400)]
2752: Move arv-put functionality to arvados.commands.put.
This will make it easier to unit test arv-put, and make it easier for
developers to extend or customize its functionality.
Brett Smith [Thu, 22 May 2014 19:49:31 +0000 (15:49 -0400)]
2752: Move Python tests to dedicated subdirectory.
This better follows Python convention, and will let us build out our
test frameworks without cluttering the root source directories.
This change brings a small change to the way we run tests. Pass the
directory name to the discover command:
$ python -m unittest discover tests
Brett Smith [Fri, 30 May 2014 14:40:07 +0000 (10:40 -0400)]
2752: Add arvados.collections.ResumableCollectionWriter.
This is a subclass of CollectionWriter that only accepts data from the
filesystem. In exchange, it can record its own state, and resume
writing from one of those states. arv-put will use this to make the
user experience nicer if a long upload is interrupted.
Brett Smith [Tue, 20 May 2014 16:14:07 +0000 (12:14 -0400)]
2752: Implement CollectionWriter with a work queue.
This will make it easier to capture and restore state.
Brett Smith [Tue, 20 May 2014 15:29:44 +0000 (11:29 -0400)]
2752: Reuse CollectionWriter.write_file() where possible.
Brett Smith [Tue, 20 May 2014 15:15:00 +0000 (11:15 -0400)]
2752: Add method CollectionWriter.write_file().
Brett Smith [Tue, 20 May 2014 15:12:56 +0000 (11:12 -0400)]
2752: Remove trailing whitespace in arv-put.
Brett Smith [Tue, 20 May 2014 14:25:18 +0000 (10:25 -0400)]
2752: Add tests for CollectionWriter.write_directory_tree().
Brett Smith [Tue, 20 May 2014 13:47:27 +0000 (09:47 -0400)]
2752: Clean up after Python SDK Collections tests.
Brett Smith [Tue, 20 May 2014 13:40:53 +0000 (09:40 -0400)]
2752: Refactor Python SDK test_collections.
This eliminates duplication in setup and teardown code, and clarifies
dependencies between tests a bit.
Brett Smith [Mon, 19 May 2014 18:10:41 +0000 (14:10 -0400)]
2752: Clean up arv-put imports.
Tom Clegg [Thu, 29 May 2014 20:48:30 +0000 (16:48 -0400)]
2762: Do not try to auto-destroy referring links when destroying a model with no uuid. refs #2762
Tom Clegg [Thu, 29 May 2014 17:58:02 +0000 (13:58 -0400)]
2754: Change "create pipeline" label to "run", to match others in same workflow.
Peter Amstutz [Thu, 29 May 2014 17:41:23 +0000 (13:41 -0400)]
Another fix to sharing scope to use keep_services/accessible. refs #2766
Peter Amstutz [Thu, 29 May 2014 17:05:43 +0000 (13:05 -0400)]
Added proper sharing scope to permit actually downloading shared files. refs #2766
Brett Smith [Thu, 29 May 2014 14:52:04 +0000 (10:52 -0400)]
2753: Fix Collection inline file view.
The height limit needs to be applied to the images themselves, rather
than the containing div.
Peter Amstutz [Thu, 29 May 2014 14:18:53 +0000 (10:18 -0400)]
Added rescue AccessForbiddenException to collections controller and view to
disable the "sharing link" feature when the client is untrusted. refs #2766
Peter Amstutz [Thu, 29 May 2014 13:32:08 +0000 (09:32 -0400)]
Merge branch '2044-share-button' refs #2766
Peter Amstutz [Thu, 29 May 2014 13:22:43 +0000 (09:22 -0400)]
Go Keep client correctly closes response body on client requests, should fix
lingering connections problem. Also added graceful shutdown on SIGINT to
keepproxy. refs #1885
radhika [Wed, 28 May 2014 16:58:11 +0000 (12:58 -0400)]
Merge branch 'master' of git.curoverse.com:arvados
radhika [Wed, 28 May 2014 16:56:36 +0000 (12:56 -0400)]
An equivalent of this script is now at services/api/script. This was done as part of Feature #2272.
Tom Clegg [Wed, 28 May 2014 16:44:04 +0000 (12:44 -0400)]
2754: Merge branch '2754-pipeline-template-description' refs #2754
Tom Clegg [Wed, 28 May 2014 16:00:33 +0000 (12:00 -0400)]
2754: Add migration missing from commit:
fe84c84827af6092784408629e4cde0663fe6af8
Peter Amstutz [Wed, 28 May 2014 15:19:37 +0000 (11:19 -0400)]
2044: Fixed CollectionsController#search_scopes to use 'filter' correctly
instead of using 'where' incorrectly. Reverted most changes to
ArvadosModel#ensure_owner_uuid_is_permitted. Added rescue to
collections/show.html.erb when log.event_at is nil.
Ward Vandewege [Wed, 28 May 2014 15:00:07 +0000 (11:00 -0400)]
Do not use symbol keys in the properties of permission links.
Peter Amstutz [Wed, 28 May 2014 14:53:23 +0000 (10:53 -0400)]
Merge branch 'master' into 2044-share-button
Tom Clegg [Wed, 28 May 2014 14:04:55 +0000 (10:04 -0400)]
2892: Fix user agreement routing. closes #2892
Tom Clegg [Wed, 28 May 2014 00:01:22 +0000 (20:01 -0400)]
2862: Use ?format=js so browser does not think js and html views are
interchangeable. closes #2862
Tom Clegg [Tue, 27 May 2014 22:16:39 +0000 (18:16 -0400)]
2760: Merge branch '2760-not-all-objects-in-folders' refs #2760
Peter Amstutz [Tue, 27 May 2014 21:07:40 +0000 (17:07 -0400)]
Merge branch 'master' of git.curoverse.com:arvados
Peter Amstutz [Tue, 27 May 2014 21:07:11 +0000 (17:07 -0400)]
Adjusted timeout on websocket_test from 3 seconds to 4 in an attempt to reduce
the number of spurious timeouts on jenkins.
radhika [Tue, 27 May 2014 20:50:26 +0000 (16:50 -0400)]
Merge branch 'master' of git.curoverse.com:arvados
Peter Amstutz [Tue, 27 May 2014 20:50:09 +0000 (16:50 -0400)]
Merge branch 'master' of git.curoverse.com:arvados #2751
radhika [Tue, 27 May 2014 20:20:57 +0000 (16:20 -0400)]
Merge branch '2756-eventbus-in-workbench' into 2678-owned_by
Conflicts:
services/api/script/crunch-dispatch.rb
radhika [Tue, 27 May 2014 19:54:36 +0000 (15:54 -0400)]
2756: fix error where partial lines were potentially logged
Peter Amstutz [Tue, 27 May 2014 19:50:04 +0000 (15:50 -0400)]
Merge branch 'master' of git.curoverse.com:arvados refs #2751
Peter Amstutz [Tue, 27 May 2014 19:49:37 +0000 (15:49 -0400)]
Merge branch '2751-python-sdk-keep-proxy-support' closes #2751
radhika [Tue, 27 May 2014 19:16:05 +0000 (15:16 -0400)]
Merge branch 'master' into 2756-eventbus-in-workbench
radhika [Tue, 27 May 2014 19:15:26 +0000 (15:15 -0400)]
2756: log span title
radhika [Tue, 27 May 2014 19:11:07 +0000 (15:11 -0400)]
2756: fix typo in helper method signature
radhika [Tue, 27 May 2014 18:59:55 +0000 (14:59 -0400)]
2756: update the ui to show job progress, and links in one column
radhika [Tue, 27 May 2014 17:19:53 +0000 (13:19 -0400)]
2525: add some javadoc comments
radhika [Tue, 27 May 2014 17:02:47 +0000 (13:02 -0400)]
2678: upon Tom's feedback, setting owner_uuid of the pi whether or not the pi belongs to a group.
Tom Clegg [Tue, 27 May 2014 15:43:07 +0000 (11:43 -0400)]
2754: Add pipeline template description attribute.
radhika [Tue, 27 May 2014 15:30:44 +0000 (11:30 -0400)]
Merge branch 'master' into 2525-java-sdk
radhika [Tue, 27 May 2014 15:22:00 +0000 (11:22 -0400)]
Merge branch 'master' into 2678-owned_by
radhika [Tue, 27 May 2014 15:16:22 +0000 (11:16 -0400)]
Merge branch 'master' into 2756-eventbus-in-workbench
radhika [Tue, 27 May 2014 15:15:03 +0000 (11:15 -0400)]
Merge branch 'master' into 2525-java-sdk
radhika [Tue, 27 May 2014 15:12:25 +0000 (11:12 -0400)]
Merge branch 'master' into 2767-doc-updates
Tom Clegg [Tue, 27 May 2014 15:08:25 +0000 (11:08 -0400)]
2760: Exclude system objects like VMs and repos from folder behavior.
Tim Pierce [Tue, 27 May 2014 15:06:42 +0000 (11:06 -0400)]
Merge branch '2755-python-sdk-permissions'
Refs #2755.
radhika [Tue, 27 May 2014 14:47:17 +0000 (10:47 -0400)]
Merge branch 'master' into 2525-java-sdk
Tom Clegg [Tue, 27 May 2014 14:44:56 +0000 (10:44 -0400)]
2760: Merge branch '2760-folder-hierarchy' refs #2760
Peter Amstutz [Tue, 27 May 2014 14:43:04 +0000 (10:43 -0400)]
2751: Test tweak to clear settings after changing environment variables.
Tim Pierce [Tue, 27 May 2014 14:38:29 +0000 (10:38 -0400)]
2865: collect garbage after each GET and PUT.
Quick fix for Keep OOM errors: reclaim memory aggressively.
Fixes #2865.
radhika [Tue, 27 May 2014 14:38:32 +0000 (10:38 -0400)]
Merge branch '2756-eventbus-in-workbench' into 2678-owned_by
radhika [Tue, 27 May 2014 14:37:57 +0000 (10:37 -0400)]
Merge branch 'master' into 2756-eventbus-in-workbench
radhika [Tue, 27 May 2014 14:37:12 +0000 (10:37 -0400)]
Merge branch 'master' into 2678-owned_by
Peter Amstutz [Tue, 27 May 2014 14:36:56 +0000 (10:36 -0400)]
2751: Changed to "except:" to "except Exception:". Refactored environment setup in keep_client_Test to be a bit more robust.
radhika [Tue, 27 May 2014 14:35:03 +0000 (10:35 -0400)]
2678: update a-r-p-i to set the owner_uuid on the newly created job when the pi belongs to a group. also, update crunch-dspatch to create the log lines with the group's owner_uuid.
Peter Amstutz [Tue, 27 May 2014 14:23:12 +0000 (10:23 -0400)]
Merge branch 'master' of git.curoverse.com:arvados refs #1885
Peter Amstutz [Tue, 27 May 2014 14:22:38 +0000 (10:22 -0400)]
1885: Tests and bugfixes for -no-get and -no-put options.
Tom Clegg [Tue, 27 May 2014 14:17:33 +0000 (10:17 -0400)]
2760: Merge branch 'master' into 2760-folder-hierarchy
radhika [Mon, 26 May 2014 17:33:45 +0000 (13:33 -0400)]
Merge branch 'master' into 2756-eventbus-in-workbench
Brett Smith [Mon, 26 May 2014 15:36:11 +0000 (11:36 -0400)]
2863: Remove debugging print from Python KeepClient.
Other tools like crunch-job parse the output of tools like arv-get and
arv-put. Right now crunch-job is saving this debug output in the log
attribute of running jobs, which confuses several other tools like
Workbench.
Closes #2863.
radhika [Sun, 25 May 2014 04:08:15 +0000 (00:08 -0400)]
2756: added a link to job when a job is chosen
radhika [Sat, 24 May 2014 22:23:44 +0000 (18:23 -0400)]
2756: log write
radhika [Sat, 24 May 2014 21:08:06 +0000 (17:08 -0400)]
2756: fix issue from merge conflict