* Import addons-nginx into addons-server
* Serve xpis through X-Accel-Redirect instead of redirecting to dedicated domain
The so called "CDN URLs" for files are less useful now that even AMO main domain
is behind a CDN, and we want to move away from it because internally it exposes
paths on EFS that nginx is serving directly, making directory structure changes
difficult.
This commit focuses on the download_file/download_latest endpoints
addons-server exposes, and leaves the old dedicated CDN domain intact for the
moment - it is still used by the update service but will go away in a future
commit.
* Few fixes and test updates
* filename is an optional argument
* Test fixes
* Add assertions about caching
* Fix latest url
* Rename type_ to download_type
* reformat url definition
We're using MySQL 8.0 from MySQL, not MariaDB, so client libs should
follow for consistency - especially as MariaDB doesn't support the
default auth method MySQL 8 uses.
Also contains a slight refactor of the Dockerfile-s to have fewer
steps and be more consistent with each other.
* Move to a rolling debian-testing release for libmagic and libgit2
Fixes#13487
Unblocks #13391
And at least refs #12924, #12357 (these need further testing)
* Roll back temporary change
* Move CI envvar definition to selenium override config.
This should fix autoreload.
* Actually use to be more explicit
* Disable integration tests
* Try out different approach
* Add docs, move back the variable definition to docker-compose.yml
* Update comment
* Fix uwsgi autoreload.
This will scan python modules for modifications every 3 seconds. It most
certainly will still be more efficient than Django's autoreload.
Fixes#11990
* re-add vaccum
* Use watchmedo to reload properly
* Update 'file' version to fix json<->html detection for reviewer api.
Fixes#11534
* Add fixture file
* Rollback to xenial, install libmagic-dev manually
* Add libmagic1 dependency too
* test amd64
* Add required libmagic-mgc dependency
* Add explaining comment
* Try out using ubuntu disco sources for libmagic installation
* Force installation from the disco repo.
* Add debian stretch backports repo as discussed with @autrilla
* Add libgit2-dev dependency that's needed by pygit2 (via debian stretch
to get the proper version that's required by up-to-date pygit2)
* Add pygit 0.27.1
Fixes#9496
* Make also use of setting PYTHONDONTWRITEBYTECODE inside the docker
image to reduce it's size considerably
* Remove some outdated code-parts
* Rename `olympia` inside our worker image to `olympia-worker`
* remove all build and cache folders after every `make update_deps`
(which might help avoid any more package update/install ambiquities)
Fixes#7518
Revert "docker: Fix Debian build issue for python package"
This reverts commit 2239e8483e.
Revert "docker: Include uwsgi plugin"
This reverts commit 807420bf16.
Revert "docker: Include uwsgi"
This reverts commit ceb36b5890.
Revert "docker: Update deploy to use Docker Python image"
This reverts commit 916547d2b5.
Revert "docker: Remove ‘centos’ from docker-indicating file"
This reverts commit d34b92e0e8.
Revert "docker: Update build to use Docker Python image"
This reverts commit b08b51375d.
This provides an updated Python (2.7.14) as part of
https://bugzilla.mozilla.org/show_bug.cgi?id=1390789
The Docker images are not based on Centos so package
names needed to be updated. Comparable packages were
added, but may not contain the same versions or even
components.
* Update celery, config and dependencies to 4.x
* Remove usage of TaskSet, replace with group(), update `ampq`
* Enable redis on travis for better celery 4.x testing
* Fix lib/celery_tests
* Correct test testing email sending retrying on errors
* Fix celery subtask mocking errors.
* Fix celery broker url env variable
* Integrate watchdog for celery autoreload
* Ensure dependencies are updated for worker and web container.
* Restart all services after the dependency update
* Remove explicit celery loader environment variable.
* Restart all services in web and worker after running 'update_docker'
* Increase sleep time a bit
This reverts commit 541688d5f2.
This caused some breakage on dev. We suspect it's because newrelic
hooks into the celery task launching code but does not support
celery 4.x yet.
* Update celery config to seamlessly upgrade to 4.x
* Update celery initialization for new configs and 4.x
* Update dependencies to celery 4.x
* Fix flake8
* Fix typo
* Remove usage of TaskSet, replace with group(), update `ampq`
* Enable redis on travis for better celery 4.x testing
* Fix more test setting names
* Fix lib/celery_tests
* Correct test testing email sending retrying on errors
* Fix celery subtask mocking errors.
* Fix new line
* Fix celery broker url env variable
* Integrate watchdog for celery autoreload
* Try something weird... ui-tests
* do we need extra hosts?
* more sleep?
* Try logging, try docker run to avoid port collisions
* More debugging
* Cleanup again
* Fix logs?
* Ensure dependencies are updated for worker and web container.
* Remove not functioning log saving for onw
* Restart all services after the dependency update
* Revert "Remove old node 4 repo"
This reverts commit 7d221f0814.
* Add back nodesource v6 repo
We need to use nodesource v6 repo for latest security fixes. EPEL updates
are often delayed.
* Upgrade to Centos 7
* Use gpg-keys for nodesource and mysql56 repositories
* Uses more prod-like package versions
* Update pyrepo links to use https://pyrepo.stage.mozaws.net/olympia/
(this is what we're using on prod)