Merge lp:~robru/bileto/enable-britney into lp:bileto
| Status: | Merged |
|---|---|
| Approved by: | Robert Bruce Park on 2015-12-14 |
| Approved revision: | 431 |
| Merged at revision: | 332 |
| Proposed branch: | lp:~robru/bileto/enable-britney |
| Merge into: | lp:bileto |
| Diff against target: |
727 lines (+435/-39) 16 files modified
.bzr-builddeb/default.conf (+2/-0) Makefile (+11/-2) britney/Makefile (+6/-0) britney/britney.conf.in (+58/-0) britney/cron (+2/-0) britney/expire.sh (+8/-0) britney/fetch-indexes (+139/-0) britney/iterate.py (+150/-0) db_migrations.sh (+27/-23) debian/control (+1/-0) tests/test_v1.py (+8/-3) tickets/models.py (+6/-2) tickets/settings.py (+2/-0) tickets/static/app.js (+4/-4) tickets/static/index.html (+6/-1) tickets/v1.py (+5/-4) |
| To merge this branch: | bzr merge lp:~robru/bileto/enable-britney |
| Related bugs: |
| Reviewer | Review Type | Date Requested | Status |
|---|---|---|---|
| Martin Pitt | Approve on 2015-12-14 | ||
| Robert Bruce Park (community) | Approve on 2015-12-14 | ||
|
Review via email:
|
|||
Commit Message
Run britney inside bileto.
Description of the Change
Experimental addition of britney to bileto instance.
| Robert Bruce Park (robru) wrote : | # |
| Robert Bruce Park (robru) wrote : | # |
Ok, this is looking great for a first iteration:
* All requests in qa state 'Ready for QA' are automatically submitted for autopkgtests.
* Results automatically appear in the request page, with links to the excuses page.
* Publication is not blocked on this being successful
* The huge archive indexes that britney consumes are downloaded in parallel
* each britney is run in series due to incredible memory consumption (I'll be increasing the production unit from 2GB to 8GB when this goes live)
| Robert Bruce Park (robru) wrote : | # |
Here is the successful test run:
https:/
(note to self, make train post this on MRs)
| Robert Bruce Park (robru) wrote : | # |
Also here's the staging ticket with the links to the excuses pages:
https:/
And the log:
https:/
| Martin Pitt (pitti) wrote : | # |
> * each britney is run in series due to incredible memory consumption (I'll be increasing the production unit from 2GB to 8GB when this goes live)
Yes, it does that I'm afraid -- it slurps in the entire Packages_*/Sources indexes as Python dictionaries so that it can make efficient installability checks, reverse and forward dependency calculations, etc. So its RAM usage can't be significantly reduced.
Why does the ticket say "Ready for QA" while tests are still running, and thus packages are still "not considered"? That smells like a bug in evaluating excuses.yaml?
| Martin Pitt (pitti) wrote : | # |
A few nits, but by and large LGTM. I suggest adding s390x in a separate MP to not make this too messy.
- 432. By Robert Bruce Park on 2015-12-14
-
Permanently enable verbose output for now.
- 433. By Robert Bruce Park on 2015-12-14
-
Fix amqp uri leak.
| Robert Bruce Park (robru) wrote : | # |
some replies inline.
- 434. By Robert Bruce Park on 2015-12-14
-
Record a log of bzr update.
- 435. By Robert Bruce Park on 2015-12-14
-
Run autopkgtests on s390x also.
- 436. By Robert Bruce Park on 2015-12-14
-
Stupid bzr logging on stderr.
- 437. By Robert Bruce Park on 2015-12-14
-
Use series-specific configs with different arches.
- 438. By Robert Bruce Park on 2015-12-14
-
Hard-code s390x in ADT_ARCHES as it is ignored pre-xenial.
- 439. By Robert Bruce Park on 2015-12-14
-
Fix s390x in xenial.
- 440. By Robert Bruce Park on 2015-12-14
-
Drop back to every 15 minutes in cron.

This is so close to done I can smell it.
TODO:
* Expose britney output in flask
* Figure out why vivid doesn't work since xenial is working, and also figure out why the logging shows some things but not python tracebacks. I guess STDERR is getting lost somewhere along the way.