- Adds a new "infrastructure" section to the docs, which describes
architecture, administration and troubleshooting (fixes bug 1165259).
- Adds code comments to any deployment-related files in the repository.
- Adds documentation for the various ways in which users can access
Treeherder data (fixes bug 1335172).
- Reorganises the structure of some of the existing non-infrastructure
docs, to make the documentation easier to navigate.
Since they are out of sync with `Procfile` and long term are best replaced
with one of the solutions in bug 1318020.
The development setup instructions have been updated to stop recommending
gunicorn, since Django's runserver is more appropriate for beginners
(and I'm pretty sure the memory leaks seen previously no longer occur).
The latest policy used in the report-only header has been working well
on production (the violation reports logged to New Relic are only from
scripts injected by browser addons), so we're ready to start enforcing
the policy by using the real `Content-Security-Policy` header name.
NB: When features are added in the future, PR authors and reviewers will
need to remember to update the policy if needed (for example to add domains
to the `connect-src` directive). The CSP header is not enabled when using
`webpack-dev-server` (it would break dev source maps and react-hot-loader)
so if in doubt test locally (using `yarn build` and serving via Django
runserver) or on prototype first.
See:
https://developer.mozilla.org/en-US/docs/Web/HTTP/CSPhttps://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Content-Security-Policy
To ensure the CI and local environments match stage/production (where
Elasticsearch is not configured) and don't run differing code-paths due
to the various `ELASTICSEARCH_URL` conditionals used in auto-classify.
At such time when work on the fuzzy auto-classify matcher continues,
this can be reverted.
Also remove all Karma support and update the docs to only mention ``Jest``.
One of the test files was testing some AngularJS filters. I converted these
tests to test the equivalent helper functions.
Since it's only used by the frontend as a fallback for when it cannot
find a non-gzipped version of `runnable-jobs.json`, and enough time has
now passed for all jobs to have that file.
Switches from Sphinx to MkDocs, since it:
* supports Markdown natively without requiring hacks like `eval_rst`
* validates inline links, ensuring that they are not broken
* has a more pleasant live-reloading dev-server
* supports the nicer looking mkdocs-material theme
* is a third of the size of Sphinx (including deps)
The theme change is now possible since Read the Docs have just started
supporting use of custom themes (previously they would override the theme
and use `readthedocs` theme regardless).
Currently Prettier is used via the ESLint integration, which only supports
JS/JSX and not the other filetypes Prettier is able to format.
For now these additional filetypes are excluded by `.prettierignore`
entries, however these will be removed/fixes applied in later commits.
So that their purposes are clearer and it's easier to differentiate
between generated content and files committed to the repository.
* Neutrino build: `build/` -> `.build/`
* Sphinx build: `_build/` -> `.build-docs/`
* Django collectstatic: `treeherder/static/` -> `.django-static/`
Since it's more reliable (and strict) at code formatting than ESLint.
We use it via an ESLint plugin, and so disable the style-related AirBnB
preset rules, leaving the AirBnB guide to handle only correctness and
best practices rules.
It's highly encouraged to use an IDE integration or Git commit hook
to run Prettier (or `yarn lint --fix`) automatically. See:
* https://prettier.io/docs/en/editors.html
* https://prettier.io/docs/en/precommit.html
We may consider enabling a git commit hook out of the box (using eg
Husky) in the future, however they have previously been known to
interfere with partial-staging workflows, so would need to test the
fixes they have made for them thoroughly first.
In future PRs we may also want to start formatting JSON/CSS/Markdown
using Prettier too.
Imports must now be grouped like so (with newlines between each):
```
// "external" modules
import _ from 'lodash';
import chalk from 'chalk';
// modules from a "parent" directory
import foo from '../foo';
import qux from '../../foo/qux';
// "sibling" modules from the same or a sibling's directory
import bar from './bar';
import baz from './bar/baz';
```
The `import/order` rule has auto-fix support, so any errors can be
resolved using `yarn lint --fix`.
See:
https://github.com/benmosher/eslint-plugin-import/blob/master/docs/rules/order.md
Neutrino controls our frontend linting, transpilation, source-maps,
testing, dev-server and optimisation of production builds.
Highlights of the upgrade are:
* Major version updates to the individual tools within (such as webpack,
Babel and ESLint), significantly improving performance, fixing
transpilation/minification correctness bugs, adding support for newer
ECMAScript features, and increasing linter coverage.
* Hot reloading in the dev server now works for all entry-points and not
just the jobs view, shortening the feedback cycle.
* Reduced bundle size due to webpack 4's tree shaking, scope hoisting,
automatic shared/vendor code chunk splitting (no need for the manually
maintained 'vendor' list).
* CSS is now extracted out of JS, which improves performance, reduces
bundle size and prevents the initial white flash of un-styled content.
* Support for dynamic imports/code splitting (needed for bug 1502192).
* Support for Jest via a new Jest preset (unblocks bug 1364045).
* Support for public class field declarations (unblocks bug 1480166).
* Improved source-maps (increases the quality of production exception
trace-backs and fixes several debugger breakpoint bugs).
* Reduced amount of custom configuration required for our fairly complex
frontend needs, reducing maintenance burden and allowing for easier
future Neutrino upgrades.
In addition this PR:
* Fixes the WhiteNoise `immutable_file_test()` regex, so that it now
correctly enables browser caching of images, fonts and source maps.
* Enables webpack-dev-server's overlay feature, which displays any
compilation errors in the browser, saving having to switch back
to the console (this can be enabled for warnings too if desired).
* Enables webpack-dev-server's automatic browser-opening feature,
which saves having to manually navigate to `localhost:5000` after
running `yarn start`.
* Switches Karma tests to run Firefox in headless mode, reducing the
workflow disruption when running `yarn test`.
* Uses the new webpack `performance` option to enable maximum asset
file size thresholds, to help prevent bundle-size regressions.
* Rewrites the `package.json` script commands so that they now work
correctly on Windows, even when setting environment variables.
Performance comparison:
* Local `yarn build`:
- Cached: 2m34s -> 23s
- Uncached: 2m34s -> 58s
* Local `yarn start`:
- Cached: 34.5s -> 13.6s
- Uncached: 34.5s -> 31.3s
* Local `yarn test`
- Cached: 61.5s -> 19.8s
- Uncached: 61.5s -> 22.0s
* Local `yarn lint`
- Cached: 3.8s -> 1.8s
- Uncached: 13.7s -> 13.4s
* Travis end-to-end time:
9 minutes -> 6 minutes
* Heroku deploy end-to-end time:
14 minutes -> 9 minutes
* Enables the display of skipped test/expected fail reasons, in
the pytest summary.
* Skips the Selenium tests with a clear reason message, unless the
built UI is found (preventing the annoying/confusing test timeouts).
* Removes the disabling of the `pytest-html` and `pytest-metadata`
plugins, since they are required when passing the `--html` option
to generate an HTML report.
* Updates the docs to mention `yarn build` and `--html`.
Now that autophone and AWFY have migrated to Taskcluster, there are
no more submitters of jobs to the REST API (confirmed via New Relic
Insights). As such, this deprecated data ingestion method can now be
removed, along with support for API Hawk auth, API POST throttling
and `treeherder-client` job submission capability.
After this lands we'll need to manually drop the `credentials` table.
Runnable jobs for buildbot were calculated via a celerybeat task
(that was disabled in #4007) and the results stored in the
`runnable_jobs` table. This can all be removed now that buildbot is
EOL, since the remaining support for Taskcluster runnable jobs does
not use that celery task/Django model.
We don't currently change the push ingestion sources in any of our envs
(prod, staging, prototype, test, dev) from their defaults. This
hardcodes those unchanging values and documents where to change them
should an engineer need to.
This reverts commit 1728dd306c (#3884), due to:
```
Pulse warning: queue "queue/treeherder-staging/jobs" is overgrowing
Warning: your queue "queue/treeherder-staging/jobs" on exchange "could not be determined" is
overgrowing (4083 ready messages, 4083 total messages).
...
```
Renaming as a separate commit since otherwise Git was struggling to
identify the renames, making the diff harder to review.
`index.rst` has been left unchanged, since we still want to use the
`toctree` directive for now.
We want to switch from reStructuredText to Markdown, since:
* it's more user-friendly
* it is used in multiple places on GitHub, so people are more familiar
with it / context switching between the two is annoying
* the disadvantages of markdown typically raised as reasons to prefer
reST either don't apply to our use-case, or are no longer accurate
with recent tooling/extensions
* switching means we can also later start using mkdocs (which doesn't
support reST) instead of Sphinx. That change has been kept separate
to reduce the size of the PR, and to make the comparison between
the two more fair.
See:
https://docs.readthedocs.io/en/latest/getting_started.html#in-markdownhttps://recommonmark.readthedocs.io/en/latest/index.html
Previously the mixture of heading types was confusing the RTD theme,
causing the navigation sidebar to display at the top level not just
the page links, but links to subsections within those pages too.
The subsection names taken out of context were pretty confusing,
making it hard to navigate the docs.
The architecture pages are significantly out of date and so have
been removed until such time as new ones are created (this will be
simpler once Buildbot/REST API support is removed later this year).
In addition, the deployment page duplicated the UI installation page.
The `--provision` flag is unnecessary for new instances, since Vagrant
runs `provision` as part of the initial `up` regardless. However by
including `--provision` in the suggested command, it will mean
returning contributors will re-run provision when they re-use the
same steps.
An alternative approach would be to mark the provision block in
`Vagrantfile` as `run: "always"`, however that would slow down the
`vagrant up` of core developers too (by 40s), with no way to opt-out.
Previously it was not possible to test features that required an
authenticated user when:
* using `yarn start` with Vagrant (bug 1363722), which meant slower
watch builds
* pointing the UI at the prod/stage API (bug 1317752), which was
extremely limiting
Now login works in all environments, since the frontend no longer uses a
URL prefix, but instead webpack-dev-server proxies non-webpack URLs to
the chosen `BACKEND_DOMAIN` - avoiding cross-domain issues. Cookies are
rewritten to remove any `secure` directive (which is set on production),
so that they can still be read from HTTP localhost. The `Referer` has to
also be changed to stop Django's CSRF checks from rejecting request.
The slower "build into `dist` and watch" mode is therefore no longer
necessary, so `yarn start:local` instead invokes webpack-dev-server just
like `yarn start` - and the `local-watch.js` workaround has been
removed.
Support for the "publish to GitHub with hardcoded `SERVICE_DOMAIN`"
workflow has been dropped, since it was already rarely used and there is
no way to make it support login.
The API domain environment variable was renamed to `BACKEND_DOMAIN` to
avoid potential confusion given it no longer behaves the same as
`SERVICE_DOMAIN` used to.
NB: For full stack Vagrant workflows users must now connect to port
*5000* on localhost, not 8000.
This helps prevent:
https://www.owasp.org/index.php/Reverse_Tabnabbing
We're not also using `noreferrer`, since most browsers now support
`noopener` (https://caniuse.com/#search=noopener) and the link targets
are all Mozilla properties where the referrer may be useful.
The auth.js `window.open()` has not been changed, since the login
callback makes use of `window.opener`.
With ES6, the `'use strict'` directives are unnecessary:
https://eslint.org/docs/rules/strict
The directives have been left in the Neutrino configs, since they
are used by node directly, which doesn't yet support ES6 modules.
Now that we're using MySQL 5.7, we can specify `REQUIRE SSL` on the
`CREATE USER` statement, rather than having to do so on the individual
GRANTs. Compare:
https://dev.mysql.com/doc/refman/5.6/en/create-user.htmlhttps://dev.mysql.com/doc/refman/5.7/en/create-user.html
Prevents:
```
1 warning(s): 1287 Using GRANT statement to modify existing user's
properties other than privileges is deprecated and will be removed
in future release. Use ALTER USER statement for this operation.
```
Generated using the approach documented at the end of the page:
https://treeherder.readthedocs.io/admin.html#direct-database-access
The changes are required since bug 1373008 added the `group` and
`group_failure_lines` tables and #2532 removed `text_log_summary`
and `text_log_summary_line`.
The UI has already been removed. This cleans up the data ingestion
and removes the `JobDuration` model, however leaves the `running_eta`
field on the `Job` model for the next time that table is touched (since
the table is large, so altering the schema would likely require
downtime).
Most of the existing config (generated by `sphinx-quickstart`) was
either commented out, identical to the defaults or else for build
modes we do not use - and only end up making the config hard to read.
* Removes the unnecessary initial `make html`, since sphinx-autobuild
performs the initial build itself.
* Enables `--poll` watch mode, since file change detection otherwise
doesn't work in Vagrant.
* Updates the docs to recommend the livereload web server over the
one-off build when working on docs locally.
The build previously contained warnings like:
```
docs/rest_api.rst:144: WARNING: Title underline too short.
docs/admin.rst:18: WARNING: Could not lex literal_block as "sql". Highlighting skipped.
WARNING: html_static_path entry u'.../docs/_static' does not exist
```
Now that no submissions are using revision_hash, it can be removed.
This removes everything but the model field, which will be handled
later.
I've removed revision_hash from the Pulse jobs schema without bumping
the version, which wouldn't normally be ok, but no one is still using
it, and I'd rather have explicit failures later than if we left the
schema unchanged.
Since for people installing node/yarn on their own (outside the
Vagrant environment) it can be confusing to work out which version is
appropriate, given node always has both an LTS and current release.
Note: This leaves the models and tables intact so that we can
revert without data loss in case we discover an issue. A follow-up
commit will remove those tables and models.
The maintenance section was not addressing that data can take a while before the different scheduled processes bring the data into the right tables.
This addresses documenting how to:
* Update the runnable jobs table
* Update the job priority table
This populates the job priority table with data locally.
It puts all jobs into the table without considerations of analyzing failures.
That can follow up in the future.
Update docs accordingly.
Change of new environment variable `PULSE_PUSH_SOURCES`.
Keep old `publish-resultset-runnable-job-action` task name by creating a
method that points to `publish_push_runnable_job_action`.
By default webservers like Django's runserver, gunicorn or the
Webpack devserver only bind to the loopback adapter (127.0.0.1) and
so are not accessible from outside the Vagrant / virtualbox VM,
since port forwarding only forwards traffic to the non-loopback
adapters.
Previously varnish (which listened on `0.0.0.0`) was reverse
proxying traffic to runserver/gunicorn, however we need to now do so
for webpack-dev-server on another port too. Doing both with varnish
adds complexity, and we don't actually need any of varnish's other
features, so ideally want to stop using it.
Rather than having to override each webserver to bind to all
adapters (using the IP `0.0.0.0`), it's possible to forward traffic
to the loopback adapter using iptables NAT PREROUTING rules. This
is still secure so long as the Vagrantfile port forwarding uses a
`host_ip` of `127.0.0.1`. To prevent this "Martian packet" traffic
from being blocked, `route_localnet` must also be set to `1`. See:
https://unix.stackexchange.com/questions/111433/iptables-redirect-outside-requests-to-127-0-0-1
By default neither sysctl or iptables settings are persisted across
reboots, and fixing that requires more complexity (eg installing the
iptables-persistent package and handling config changes during
provision). As such, it's just easier to re-run the commands on each
login since they take <30ms.
Previously a forgotten-about `local.conf.js` (which is git-ignored)
would override the URL passed by the `SERVICE_URL` environment variable.
With webpack and environment variables, there is no need to use a local
config file to control the API URL, so we can now remove this footgun.
Artifacts no longer exist (they've been replaced by more specific types
like "jobdetails"), and so fetching from this endpoint has been disabled
for some time already.
For data submission, we still call them artifacts (and sort their type
after submission), however all artifacts are currently submitted at the
same time as the job, so this endpoint is unused.
This import only affects internal treeherder usage, people using the
PyPI package import from the `thclient` subdirectory instead.
Fixes:
treeherder/client/__init__.py:1:1: F401 '.thclient.*' imported but unused
Since it's faster, deterministic and doesn't given obscure errors when
using `--no-bin-links` (which is required for both npm and yarn on
Windows hosts), and as such unblocks the work in bug 1343624.
Many of the commands are the same as with npm. See:
https://yarnpkg.com/en/docs/usage
The `test` script entry in `package.json` (used by `npm test`) already
calls karma with the appropriate parameters, so the helper scripts are
unnecessary.
For the same reason as the previous commit.
Ideally we'd remove the grunt abstraction entirely and call eslint from
the `lint` command, but we might as well save that to the Neutrino PR.
Routing commands via npm/yarn is preferred, since it avoids
having to do global installs of grunt-cli, which simplifies contributor
setup, and means less effort when we switch to Yarn (since it requires
manual PATH setup for globally installed packages).
Cameron Dawson
Ed Morley
KWierso
Sarah Clements
George Hickman
George Hickman
ionutgoldan
Jonathan French
Dave Hunt
Kiki
Armen Zambrano
Armen Zambrano G
Max Chehab
William Lachance
Ed Morley
Ed Morley
Casey Williams