Add CloudOps-ready Docker builds (#7072)

* First stab at dockerflow support * Add dockerfile and entrypoint for production builds * Add docker build to circle * Add stub version.json file * Fix linting issues * Add entrypoints for cronjobs
2021-03-26 16:21:57 -04:00 · 2021-03-26 16:21:57 -04:00 · 66f1eac118
--- a/.circleci/config.yml
+++ b/.circleci/config.yml
@ -54,6 +54,62 @@ jobs:
      - run: tox -e docker  # Run tests and coverage within Docker container
      - run: codecov --required -f coverage.xml

+  test-docker-build:
+    docker:
+      - image: docker:18.02.0-ce
+    steps:
+      - setup_remote_docker
+      - checkout
+      - run:
+          name: Create a version.json
+          command: |
+            # create a version.json per https://github.com/mozilla-services/Dockerflow/blob/master/docs/version_object.md
+            printf '{"commit":"%s","version":"%s","source":"https://github.com/%s/%s","build":"%s"}\n' \
+            "$CIRCLE_SHA1" \
+            "$CIRCLE_TAG" \
+            "$CIRCLE_PROJECT_USERNAME" \
+            "$CIRCLE_PROJECT_REPONAME" \
+            "$CIRCLE_BUILD_URL" > version.json
+      - run:
+          name: Build Docker image
+          command: docker build -f docker/Dockerfile -t app:build .
+
+  deploy:
+    docker:
+      - image: docker:18.02.0-ce
+    steps:
+      - setup_remote_docker
+      - checkout
+      - run:
+          name: Create a version.json
+          command: |
+            # create a version.json per https://github.com/mozilla-services/Dockerflow/blob/master/docs/version_object.md
+            printf '{"commit":"%s","version":"%s","source":"https://github.com/%s/%s","build":"%s"}\n' \
+            "$CIRCLE_SHA1" \
+            "$CIRCLE_TAG" \
+            "$CIRCLE_PROJECT_USERNAME" \
+            "$CIRCLE_PROJECT_REPONAME" \
+            "$CIRCLE_BUILD_URL" > version.json
+      - run:
+          name: Build Docker image
+          command: docker build -f docker/Dockerfile -t app:build .
+
+      - run:
+          name: Deploy to Dockerhub
+          command: |
+            echo $DOCKER_PASS | docker login -u $DOCKER_USER --password-stdin
+            # deploy master
+            if [ "${CIRCLE_BRANCH}" == "master" ]; then
+              docker tag app:build ${DOCKERHUB_REPO}:latest
+              docker push ${DOCKERHUB_REPO}:latest
+            elif  [ ! -z "${CIRCLE_TAG}" ]; then
+            # deploy a release tag...
+              echo "${DOCKERHUB_REPO}:${CIRCLE_TAG}"
+              docker tag app:build "${DOCKERHUB_REPO}:${CIRCLE_TAG}"
+              docker images
+              docker push "${DOCKERHUB_REPO}:${CIRCLE_TAG}"
+            fi
+
 orbs:
  node: circleci/node@4.1.0
  docker: circleci/docker@1.5.0
@ -64,3 +120,10 @@ workflows:
      - javascript-tests
      - builds
      - python-tests
+      - test-docker-build
+      - deploy:
+          filters:
+            tags:
+              only: /.*/
+            branches:
+              only: master
--- a/.dockerignore
+++ b/.dockerignore
@ -1,6 +0,0 @@
-# Exclude all files but those used by the `Dockerfile` to speed up
-# the transferring of the build context during the Docker build.
-# The rest of the source will be made available via bind mounts.
-
-*
-!requirements/
--- a/docker/Dockerfile
+++ b/docker/Dockerfile
@ -0,0 +1,42 @@
+## Frontend stage
+FROM node:12.14.1 AS frontend
+
+WORKDIR /app
+
+COPY ui/ /app/ui/
+COPY .neutrinorc.js package.json webpack.config.js yarn.lock /app/
+
+RUN npm install -g --force yarn@1.22.5
+RUN yarn install
+RUN yarn build
+
+
+## Backend stage
+FROM python:3.7.10-slim
+
+# libmysqlclient-dev is required for the mysqlclient Python package.
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    default-libmysqlclient-dev \
+    && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /app
+
+COPY requirements/ /app/requirements/
+RUN apt-get update && apt-get install -q --yes gcc && \
+    pip install -r requirements/common.txt && \
+    apt-get -q --yes remove gcc && \
+    apt-get -q --yes autoremove && \
+    apt-get clean && \
+    rm -rf /root/.cache
+
+COPY bin/ /app/bin/
+COPY docker/entrypoint_prod.sh /app/docker/entrypoint_prod.sh
+COPY treeherder/ /app/treeherder/
+COPY manage.py newrelic.ini version.json /app/
+
+COPY --from=frontend /app/.build/ /app/.build/
+
+RUN python manage.py collectstatic --noinput
+
+ENTRYPOINT ["/bin/bash", "/app/docker/entrypoint_prod.sh"]
+CMD ["web"]
--- a/docker/entrypoint_prod.sh
+++ b/docker/entrypoint_prod.sh
@ -0,0 +1,99 @@
+#!/bin/bash
+
+# The `release` process type specifies the command to run during deployment, and is where
+# we run DB migrations and other tasks that are 'release' rather than 'build' specific:
+# https://devcenter.heroku.com/articles/release-phase
+# https://devcenter.heroku.com/articles/runtime-principles#build-release-run
+if [ "$1" == "release" ]; then
+    exec ../bin/pre_deploy
+
+# The `web` process type is the only one that receives external traffic from Heroku's routers.
+# We set the maximum request duration to 20 seconds, to ensure that poorly performing API
+# queries do not consume a gunicorn worker for unbounded lengths of time. See:
+# https://devcenter.heroku.com/articles/python-gunicorn
+# The Heroku Python buildpack sets some sensible gunicorn defaults via environment variables:
+# https://github.com/heroku/heroku-buildpack-python/blob/master/vendor/python.gunicorn.sh
+# https://github.com/heroku/heroku-buildpack-python/blob/master/vendor/WEB_CONCURRENCY.sh
+# TODO: Experiment with different dyno sizes and gunicorn concurrency/worker types (bug 1175472).
+elif [ "$1" == "web" ]; then
+    exec newrelic-admin run-program gunicorn treeherder.config.wsgi:application --timeout 20 --bind 0.0.0.0
+
+# All other process types can have arbitrary names.
+# The Celery options such as `--without-heartbeat` are from the recommendations here:
+# https://www.cloudamqp.com/docs/celery.html
+# The REMAP_SIGTERM is as recommended by:
+# https://devcenter.heroku.com/articles/celery-heroku#using-remap_sigterm
+
+# This schedules (but does not run itself) the cron-like tasks listed in `CELERY_BEAT_SCHEDULE`.
+# However we're moving away from using this in favour of the Heroku scheduler addon.
+# NB: This should not be scaled up to more than 1 dyno otherwise duplicate tasks will be scheduled.
+# TODO: Move the remaining tasks to the addon and remove this process type (deps of bug 1176492).
+elif [ "$1" == "celery_scheduler" ]; then
+    export REMAP_SIGTERM=SIGQUIT
+    exec newrelic-admin run-program celery beat -A treeherder
+
+# Push/job data is consumed from exchanges on pulse.mozilla.org using these kombu-powered
+# Django management commands. They do not ingest the data themselves, instead adding tasks
+# to the `store_pulse_{pushes,jobs}` queues for `worker_store_pulse_data` to process.
+# NB: These should not be scaled up to more than 1 of each.
+# TODO: Merge these two listeners into one since they use so little CPU each (bug 1530965).
+elif [ "$1" == "pulse_listener_pushes" ]; then
+    exec newrelic-admin run-program ./manage.py pulse_listener_pushes
+elif [ "$1" == "pulse_listener_tasks" ]; then
+    exec newrelic-admin run-program ./manage.py pulse_listener_tasks
+
+# Processes pushes/jobs from Pulse that were collected by `pulse_listener_{pushes,tasks}`.
+elif [ "$1" == "worker_store_pulse_data" ]; then
+    export REMAP_SIGTERM=SIGQUIT
+    exec newrelic-admin run-program celery worker -A treeherder --without-gossip --without-mingle --without-heartbeat -Q store_pulse_pushes,store_pulse_tasks --concurrency=3
+
+# Handles the log parsing tasks scheduled by `worker_store_pulse_data` as part of job ingestion.
+elif [ "$1" == "worker_log_parser" ]; then
+    export REMAP_SIGTERM=SIGQUIT
+    exec newrelic-admin run-program celery worker -A treeherder --without-gossip --without-mingle --without-heartbeat -Q log_parser --concurrency=7
+elif [ "$1" == "worker_log_parser_fail_raw_sheriffed" ]; then
+    export REMAP_SIGTERM=SIGQUIT
+    exec newrelic-admin run-program celery worker -A treeherder --without-gossip --without-mingle --without-heartbeat -Q log_parser_fail_raw_sheriffed --concurrency=1
+elif [ "$1" == "worker_log_parser_fail_raw_unsheriffed" ]; then
+    export REMAP_SIGTERM=SIGQUIT
+    exec newrelic-admin run-program celery worker -A treeherder --without-gossip --without-mingle --without-heartbeat -Q log_parser_fail_raw_unsheriffed --concurrency=1
+elif [ "$1" == "worker_log_parser_fail_json_sheriffed" ]; then
+    export REMAP_SIGTERM=SIGQUIT
+    exec newrelic-admin run-program celery worker -A treeherder --without-gossip --without-mingle --without-heartbeat -Q log_parser_fail_json_sheriffed --concurrency=7
+elif [ "$1" == "worker_log_parser_fail_json_unsheriffed" ]; then
+    export REMAP_SIGTERM=SIGQUIT
+    newrelic-admin run-program celery worker -A treeherder --without-gossip --without-mingle --without-heartbeat -Q log_parser_fail_json_unsheriffed --concurrency=7
+
+# Tasks that don't need a dedicated worker.
+elif [ "$1" == "worker_misc" ]; then
+    export REMAP_SIGTERM=SIGQUIT
+    exec newrelic-admin run-program celery worker -A treeherder --without-gossip --without-mingle --without-heartbeat -Q default,generate_perf_alerts,pushlog,seta_analyze_failures --concurrency=3
+
+# Cron jobs
+elif [ "$1" == "run_intermittents_commenter" ]; then
+    newrelic-admin run-program ./manage.py run_intermittents_commenter -m auto
+
+elif [ "$1" == "update_bugscache" ]; then
+    newrelic-admin run-program ./manage.py update_bugscache
+
+elif [ "$1" == "cache_failure_history" ]; then
+    newrelic-admin run-program ./manage.py cache_failure_history
+
+elif [ "$1" == "cycle_data" ]; then
+    CHUNK_SIZE="$2"
+    DAYS="$3"
+    # may be empty
+    FROM="$4"
+    ./manage.py cycle_data --chunk-size="${CHUNK_SIZE}" --days "${DAYS}" "${FROM}"
+
+elif [ "$1" == "perf_sheriff" ]; then
+    TIME_WINDOW="$2"
+    newrelic-admin run-program ./manage.py perf_sheriff --time-window="${TIME_WINDOW}"
+
+elif [ "$1" == "update_changelog" ]; then
+    newrelic-admin run-program ./manage.py update_changelog --days 2
+
+else
+    echo "Unknown command: $1"
+    exit 1
+fi
--- a/requirements/common.in
+++ b/requirements/common.in
@ -35,3 +35,6 @@ django-cache-memoize   # Imported as cache_memoize
 mozci

 importlib-metadata
+
+# Dockerflow/CloudOps APIs
+dockerflow
--- a/requirements/common.txt
+++ b/requirements/common.txt
@ -179,6 +179,10 @@ djangorestframework==3.12.2 \
    --hash=sha256:0209bafcb7b5010fdfec784034f059d512256424de2a0f084cb82b096d6dd6a7 \
    --hash=sha256:0898182b4737a7b584a2c73735d89816343369f259fea932d90dc78e35d8ac33
    # via -r requirements/common.in
+dockerflow==2020.10.0 \
+    --hash=sha256:226086becc436b5a1995348e26c4fb2ad1d4e5dbc10dffec0c675c9a43306c8b \
+    --hash=sha256:36787fa016e5505d71d60c36cd4e0de7b2d1e50059160bd4e93ceb62bb40d3f8
+    # via -r requirements/common.in
 ecdsa==0.14.1 \
    --hash=sha256:64c613005f13efec6541bb0a33290d0d03c27abab5f15fbab20fb0ee162bdd8e \
    --hash=sha256:e108a5fe92c67639abae3260e43561af914e7fd0d27bae6d2ec1312ae7934dfe
--- a/treeherder/config/settings.py
+++ b/treeherder/config/settings.py
@ -43,7 +43,7 @@ SITE_URL = env("SITE_URL", default='http://localhost:8000')

 SITE_HOSTNAME = furl(SITE_URL).host
 # Including localhost allows using the backend locally
-ALLOWED_HOSTS = [SITE_HOSTNAME, 'localhost']
+ALLOWED_HOSTS = [SITE_HOSTNAME, 'localhost', '127.0.0.1']

 # URL handling
 APPEND_SLASH = False
@ -70,6 +70,7 @@ INSTALLED_APPS = [
    'rest_framework',
    'corsheaders',
    'django_filters',
+    'dockerflow.django',
    # treeherder apps
    'treeherder.model',
    'treeherder.webapp',
@ -118,6 +119,7 @@ MIDDLEWARE = [
        'django.middleware.common.CommonMiddleware',
        'django.middleware.csrf.CsrfViewMiddleware',
        'django.contrib.auth.middleware.AuthenticationMiddleware',
+        'dockerflow.django.middleware.DockerflowMiddleware',
    ]
    if middleware
 ]
@ -226,9 +228,11 @@ LOGGING = {
        'standard': {
            'format': "[%(asctime)s] %(levelname)s [%(name)s:%(lineno)s] %(message)s",
        },
+        'json': {'()': 'dockerflow.logging.JsonLogFormatter', 'logger_name': 'treeherder'},
    },
    'handlers': {
        'console': {'class': 'logging.StreamHandler', 'formatter': 'standard'},
+        'json': {'class': 'logging.StreamHandler', 'formatter': 'json', 'level': 'DEBUG'},
    },
    'loggers': {
        'django': {
@ -251,6 +255,10 @@ LOGGING = {
            'handlers': ['console'],
            'level': 'WARNING',
        },
+        'request.summary': {
+            'handlers': ['json'],
+            'level': 'DEBUG',
+        },
    },
 }

@ -473,3 +481,6 @@ NOTIFY_ACCESS_TOKEN = env('NOTIFY_ACCESS_TOKEN', default=None)
 # This is only used for removing the rate limiting. You can create your own here:
 # https://github.com/settings/tokens
 GITHUB_TOKEN = env("GITHUB_TOKEN", default=None)
+
+# For dockerflow
+BASE_DIR = SRC_DIR
--- a/version.json
+++ b/version.json
@ -0,0 +1,6 @@
+{
+  "commit": "stub",
+  "version": "stub",
+  "source": "https://github.com/mozilla/treeherder",
+  "build": "https://tools.taskcluster.net/task-inspector/#XXXXXXXXXXXXXXXXXX"
+}