Joseph Schorr
bc119aed22
Clarify why we need database.UserThenDisconnect
2015-02-05 15:00:19 -05:00
Joseph Schorr
bfb0784abc
Add signing to the ACI converter
2015-02-04 15:29:24 -05:00
Jimmy Zelinskie
844a960608
linter: rm unused imports, shut the linter up
2015-02-02 16:38:58 -05:00
Joseph Schorr
84e5c0644e
Address comments
2015-02-02 14:07:32 -05:00
Joseph Schorr
9583023749
Start cleanup in prep for merge
2015-01-29 14:25:42 -05:00
Joseph Schorr
6ed28930b2
Work in progress: Docker -> ACI conversion
2015-01-13 17:46:11 -05:00
Jimmy Zelinskie
716d7a737b
Strip whitespace from ALL the things.
2014-11-24 16:07:38 -05:00
Joseph Schorr
3e550b4b71
Make sure to sort the images for the squashed image system
2014-11-13 15:13:44 -05:00
Joseph Schorr
091f821a6a
- Rename get_repo_image to get_repo_image_extended and get_repo_image_directly to get_repo_image
...
- Remove the configure call from CloseForLongOperation
- Other small fixes
2014-11-10 13:44:36 -05:00
Joseph Schorr
d5bbb57481
Change registry code to disconnect from the DB before long I/O operations
2014-11-06 18:00:52 -05:00
Joseph Schorr
23d9bd2b42
Change verbs to use short lived database connections
2014-11-06 17:50:48 -05:00
Joseph Schorr
a35bc11912
Add perf comments
2014-11-05 12:27:38 -05:00
Joseph Schorr
79e4864eb2
Move the the track_and_log code into its own module. This breaks a dependency chain between index -> common -> api -> subscribe -> common again.
2014-10-30 12:49:51 -04:00
Joseph Schorr
c1398c6d2b
- Add a log entry for repo verb handling and make the container usage calculation take it into account
...
- Move all the repo push/pull/verb logging into a central track_and_log method
- Readd images accidentally deleted in the last CL
- Make the uncompressed size migration script better handle exceptions
2014-10-29 15:42:44 -04:00
Joseph Schorr
0ef17b082b
Make sure to disconnect from the database when finished with the processes
2014-10-21 17:40:57 -04:00
Joseph Schorr
a423032e80
- Make sure we log when an exception occurs in the squashing code
...
- Have queue file always try to return any remaining data in the buffer
- Remove the raise of the exception when we get an empty tar file
2014-10-16 12:54:16 -04:00
Joseph Schorr
da28bc4ce9
- Handle missing images properly
...
- Add support for deleting directories
- Add a slew of tests for deletion of directories and other kinds of deletion and layering
2014-10-14 21:40:02 -04:00
Joseph Schorr
adc915a5eb
Fix the subprocesses to also use their own storage classes; this fixes a socket error when talking to S3
2014-10-08 16:54:03 -04:00
Joseph Schorr
d16fdde528
Fix bug in dockerloadformat and make sure we handle exceptions properly in the verb call
2014-10-08 13:43:12 -04:00
Joseph Schorr
e5055763f6
Make the squashed path smaller and handle failure cases on the curl side
2014-09-19 12:54:52 -04:00
Joseph Schorr
e273dca4b4
Change back to using a docker load format
2014-09-19 12:22:54 -04:00
Jake Moshenko
11bb8e6448
Actually store the generated image storage in the database, and allow it to be garbage collected when the parent image storage is collected.
2014-09-18 17:26:40 -04:00
Joseph Schorr
efc06b54f6
Add a TODO and some slightly better naming
2014-09-16 22:44:45 -04:00
Joseph Schorr
1cfb6fc353
Have the squashing system write the data (via a double queue system and multiprocessing) to both the client and the stack's storage. On subsequent calls, if the synthetic image exists, it will be returned directly instead of being recomputed
2014-09-16 22:43:19 -04:00
Joseph Schorr
5cca609c55
Switch back to send_file and add a bit of gzip buffering
2014-09-16 14:20:42 -04:00
Joseph Schorr
9344839295
Get squashed endpoint for docker import working
2014-09-16 11:53:54 -04:00
Joseph Schorr
e3c52fa0eb
Work in progress. This is currently broken!
2014-09-16 00:18:57 -04:00
