Dockerfile rewrite based on Ruby image with performance optimizations and size reduction, dedicated Streaming image (#26850)

Co-authored-by: “Michael <“mx@vmstan.com>
Co-authored-by: Emelia Smith <ThisIsMissEm@users.noreply.github.com>

.github/workflows/build-container-image.yml

@@ -21,6 +21,8 @@ on:
         type: string
       labels:
         type: string
+      file_to_build:
+        type: string
 
 jobs:
   build-image:
@@ -86,6 +88,7 @@ jobs:
       - uses: docker/build-push-action@v5
         with:
           context: .
+          file: ${{ inputs.file_to_build }}
           build-args: |
             MASTODON_VERSION_PRERELEASE=${{ inputs.version_prerelease }}
             MASTODON_VERSION_METADATA=${{ inputs.version_metadata }}

.github/workflows/build-nightly.yml

@@ -25,6 +25,7 @@ jobs:
     needs: compute-suffix
     uses: ./.github/workflows/build-container-image.yml
     with:
+      file_to_build: Dockerfile
       platforms: linux/amd64,linux/arm64
       use_native_arm64_builder: true
       cache: false
@@ -41,3 +42,25 @@ jobs:
         type=raw,value=nightly
         type=schedule,pattern=${{ needs.compute-suffix.outputs.prerelease }}
     secrets: inherit
+
+  build-image-streaming:
+    needs: compute-suffix
+    uses: ./.github/workflows/build-container-image.yml
+    with:
+      file_to_build: streaming/Dockerfile
+      platforms: linux/amd64,linux/arm64
+      use_native_arm64_builder: true
+      cache: false
+      push_to_images: |
+        tootsuite/mastodon-streaming
+        ghcr.io/mastodon/mastodon-streaming
+      version_prerelease: ${{ needs.compute-suffix.outputs.prerelease }}
+      labels: |
+        org.opencontainers.image.description=Nightly build image used for testing purposes
+      flavor: |
+        latest=auto
+      tags: |
+        type=raw,value=edge
+        type=raw,value=nightly
+        type=schedule,pattern=${{ needs.compute-suffix.outputs.prerelease }}
+    secrets: inherit

.github/workflows/build-push-pr.yml

@@ -29,6 +29,7 @@ jobs:
     needs: compute-suffix
     uses: ./.github/workflows/build-container-image.yml
     with:
+      file_to_build: Dockerfile
       platforms: linux/amd64,linux/arm64
       use_native_arm64_builder: true
       push_to_images: |
@@ -39,3 +40,19 @@ jobs:
       tags: |
         type=ref,event=pr
     secrets: inherit
+
+  build-image-streaming:
+    needs: compute-suffix
+    uses: ./.github/workflows/build-container-image.yml
+    with:
+      file_to_build: streaming/Dockerfile
+      platforms: linux/amd64,linux/arm64
+      use_native_arm64_builder: true
+      push_to_images: |
+        ghcr.io/mastodon/mastodon-streaming
+      version_metadata: ${{ needs.compute-suffix.outputs.metadata }}
+      flavor: |
+        latest=auto
+      tags: |
+        type=ref,event=pr
+    secrets: inherit

.github/workflows/build-releases.yml

@@ -12,6 +12,7 @@ jobs:
   build-image:
     uses: ./.github/workflows/build-container-image.yml
     with:
+      file_to_build: Dockerfile
       platforms: linux/amd64,linux/arm64
       use_native_arm64_builder: true
       push_to_images: |
@@ -27,3 +28,24 @@ jobs:
         type=pep440,pattern={{raw}}
         type=pep440,pattern=v{{major}}.{{minor}}
     secrets: inherit
+
+  build-image-streaming:
+    if: startsWith(github.ref, 'refs/tags/v4.3.')
+    uses: ./.github/workflows/build-container-image.yml
+    with:
+      file_to_build: streaming/Dockerfile
+      platforms: linux/amd64,linux/arm64
+      use_native_arm64_builder: true
+      push_to_images: |
+        tootsuite/mastodon-streaming
+        ghcr.io/mastodon/mastodon-streaming
+      # Do not use cache when building releases, so apt update is always ran and the release always contain the latest packages
+      cache: false
+      # Only tag with latest when ran against the latest stable branch
+      # This needs to be updated after each minor version release
+      flavor: |
+        latest=${{ startsWith(github.ref, 'refs/tags/v4.3.') }}
+      tags: |
+        type=pep440,pattern={{raw}}
+        type=pep440,pattern=v{{major}}.{{minor}}
+    secrets: inherit

.github/workflows/test-image-build.yml

@@ -7,6 +7,7 @@ on:
       - .github/workflows/build-releases.yml
       - .github/workflows/test-image-build.yml
       - Dockerfile
+      - streaming/Dockerfile
 permissions:
   contents: read
 
@@ -18,4 +19,17 @@ jobs:
 
     uses: ./.github/workflows/build-container-image.yml
     with:
+      file_to_build: Dockerfile
       platforms: linux/amd64 # Testing only on native platform so it is performant
+      cache: true
+
+  build-image-streaming:
+    concurrency:
+      group: ${{ github.workflow }}-${{ github.ref }}-streaming
+      cancel-in-progress: true
+
+    uses: ./.github/workflows/build-container-image.yml
+    with:
+      file_to_build: streaming/Dockerfile
+      platforms: linux/amd64 # Testing only on native platform so it is performant
+      cache: true

Dockerfile

@@ -1,20 +1,182 @@
 # syntax=docker/dockerfile:1.4
-# This needs to be bookworm-slim because the Ruby image is built on bookworm-slim
-ARG NODE_VERSION="20.9-bookworm-slim"
 
-FROM ghcr.io/moritzheiber/ruby-jemalloc:3.2.2-slim as ruby
+# Please see https://docs.docker.com/engine/reference/builder for information about
-FROM node:${NODE_VERSION} as build
+# the extended buildx capabilities used in this file.
+# Make sure multiarch TARGETPLATFORM is available for interpolation
+# See: https://docs.docker.com/build/building/multi-platform/
+ARG TARGETPLATFORM=${TARGETPLATFORM}
+ARG BUILDPLATFORM=${BUILDPLATFORM}
 
-COPY --link --from=ruby /opt/ruby /opt/ruby
+# Ruby image to use for base image, change with [--build-arg RUBY_VERSION="3.2.2"]
+ARG RUBY_VERSION="3.2.2"
+# # Node version to use in base image, change with [--build-arg NODE_MAJOR_VERSION="20"]
+ARG NODE_MAJOR_VERSION="20"
+# Debian image to use for base image, change with [--build-arg DEBIAN_VERSION="bookworm"]
+ARG DEBIAN_VERSION="bookworm"
+# Node image to use for base image based on combined variables (ex: 20-bookworm-slim)
+FROM docker.io/node:${NODE_MAJOR_VERSION}-${DEBIAN_VERSION}-slim as node
+# Ruby image to use for base image based on combined variables (ex: 3.2.2-slim-bookworm)
+FROM docker.io/ruby:${RUBY_VERSION}-slim-${DEBIAN_VERSION} as ruby
 
-ENV DEBIAN_FRONTEND="noninteractive" \
+# Resulting version string is vX.X.X-MASTODON_VERSION_PRERELEASE+MASTODON_VERSION_METADATA
-    PATH="${PATH}:/opt/ruby/bin"
+# Example: v4.2.0-nightly.2023.11.09+something
+# Overwrite existance of 'alpha.0' in version.rb [--build-arg MASTODON_VERSION_PRERELEASE="nightly.2023.11.09"]
+ARG MASTODON_VERSION_PRERELEASE=""
+# Append build metadata or fork information to version.rb [--build-arg MASTODON_VERSION_METADATA="something"]
+ARG MASTODON_VERSION_METADATA=""
 
-SHELL ["/bin/bash", "-o", "pipefail", "-c"]
+# Allow Ruby on Rails to serve static files
+# See: https://docs.joinmastodon.org/admin/config/#rails_serve_static_files
+ARG RAILS_SERVE_STATIC_FILES="true"
+# Allow to use YJIT compiler
+# See: https://github.com/ruby/ruby/blob/master/doc/yjit/yjit.md
+ARG RUBY_YJIT_ENABLE="1"
+# Timezone used by the Docker container and runtime, change with [--build-arg TZ=Europe/Berlin]
+ARG TZ="Etc/UTC"
+# Linux UID (user id) for the mastodon user, change with [--build-arg UID=1234]
+ARG UID="991"
+# Linux GID (group id) for the mastodon user, change with [--build-arg GID=1234]
+ARG GID="991"
 
+# Apply Mastodon build options based on options above
+ENV \
+# Apply Mastodon version information
+  MASTODON_VERSION_PRERELEASE="${MASTODON_VERSION_PRERELEASE}" \
+  MASTODON_VERSION_METADATA="${MASTODON_VERSION_METADATA}" \
+# Apply Mastodon static files and YJIT options
+  RAILS_SERVE_STATIC_FILES=${RAILS_SERVE_STATIC_FILES} \
+  RUBY_YJIT_ENABLE=${RUBY_YJIT_ENABLE} \
+# Apply timezone
+  TZ=${TZ}
+
+ENV \
+# Configure the IP to bind Mastodon to when serving traffic
+  BIND="0.0.0.0" \
+# Use production settings for Yarn, Node and related nodejs based tools
+  NODE_ENV="production" \
+# Use production settings for Ruby on Rails
+  RAILS_ENV="production" \
+# Add Ruby and Mastodon installation to the PATH
+  DEBIAN_FRONTEND="noninteractive" \
+  PATH="${PATH}:/opt/ruby/bin:/opt/mastodon/bin" \
+# Optimize jemalloc 5.x performance
+  MALLOC_CONF="narenas:2,background_thread:true,thp:never,dirty_decay_ms:1000,muzzy_decay_ms:0"
+
+# Set default shell used for running commands
+SHELL ["/bin/bash", "-o", "pipefail", "-o", "errexit", "-c"]
+
+ARG TARGETPLATFORM
+
+RUN echo "Target platform is $TARGETPLATFORM"
+
+RUN \
+# Sets timezone
+  echo "${TZ}" > /etc/localtime; \
+# Creates mastodon user/group and sets home directory
+  groupadd -g "${GID}" mastodon; \
+  useradd -l -u "${UID}" -g "${GID}" -m -d /opt/mastodon mastodon; \
+# Creates /mastodon symlink to /opt/mastodon
+  ln -s /opt/mastodon /mastodon;
+
+# Set /opt/mastodon as working directory
 WORKDIR /opt/mastodon
 
+# hadolint ignore=DL3008,DL3005
+RUN \
+# Mount Apt cache and lib directories from Docker buildx caches
+--mount=type=cache,id=apt-cache-${TARGETPLATFORM},target=/var/cache/apt,sharing=locked \
+--mount=type=cache,id=apt-lib-${TARGETPLATFORM},target=/var/lib/apt,sharing=locked \
+# Apt update & upgrade to check for security updates to Debian image
+  apt-get update; \
+  apt-get dist-upgrade -yq; \
+# Install jemalloc, curl and other necessary components
+  apt-get install -y --no-install-recommends \
+    ca-certificates \
+    curl \
+    ffmpeg \
+    file \
+    imagemagick \
+    libjemalloc2 \
+    patchelf \
+    procps \
+    tini \
+    tzdata \
+  ; \
+# Patch Ruby to use jemalloc
+  patchelf --add-needed libjemalloc.so.2 /usr/local/bin/ruby; \
+# Discard patchelf after use
+  apt-get purge -y \
+    patchelf \
+  ;
+
+# Create temporary build layer from base image
+FROM ruby as build
+
+# Copy Node package configuration files into working directory
+COPY package.json yarn.lock .yarnrc.yml /opt/mastodon/
+COPY .yarn /opt/mastodon/.yarn
+
+COPY --from=node /usr/local/bin /usr/local/bin
+COPY --from=node /usr/local/lib /usr/local/lib
+
+ARG TARGETPLATFORM
+
 # hadolint ignore=DL3008
+RUN \
+# Mount Apt cache and lib directories from Docker buildx caches
+--mount=type=cache,id=apt-cache-${TARGETPLATFORM},target=/var/cache/apt,sharing=locked \
+--mount=type=cache,id=apt-lib-${TARGETPLATFORM},target=/var/lib/apt,sharing=locked \
+# Install build tools and bundler dependencies from APT
+  apt-get update; \
+  apt-get install -y --no-install-recommends \
+    g++ \
+    gcc \
+    git \
+    libgdbm-dev \
+    libgmp-dev \
+    libicu-dev \
+    libidn-dev \
+    libpq-dev \
+    libssl-dev \
+    make \
+    shared-mime-info \
+    zlib1g-dev \
+  ;
+
+RUN \
+# Configure Corepack
+  rm /usr/local/bin/yarn*; \
+  corepack enable; \
+  corepack prepare --activate;
+
+# Create temporary bundler specific build layer from build layer
+FROM build as bundler
+
+ARG TARGETPLATFORM
+
+# Copy Gemfile config into working directory
+COPY Gemfile* /opt/mastodon/
+
+RUN \
+# Mount Ruby Gem caches
+--mount=type=cache,id=gem-cache-${TARGETPLATFORM},target=/usr/local/bundle/cache/,sharing=locked \
+# Configure bundle to prevent changes to Gemfile and Gemfile.lock
+  bundle config set --global frozen "true"; \
+# Configure bundle to not cache downloaded Gems
+  bundle config set --global cache_all "false"; \
+# Configure bundle to only process production Gems
+  bundle config set --local without "development test"; \
+# Configure bundle to not warn about root user
+  bundle config set silence_root_warning "true"; \
+# Download and install required Gems
+  bundle install -j"$(nproc)";
+
+# Create temporary node specific build layer from build layer
+FROM build as yarn
+
+ARG TARGETPLATFORM
+
+# Copy Node package configuration files into working directory
 RUN apt-get update && \
     apt-get -yq dist-upgrade && \
     apt-get install -y --no-install-recommends build-essential \
@@ -41,72 +203,77 @@ COPY Gemfile* package.json yarn.lock .yarnrc.yml /opt/mastodon/
 COPY streaming/package.json /opt/mastodon/streaming/
 COPY .yarn /opt/mastodon/.yarn
 
-RUN bundle install -j"$(nproc)"
+# hadolint ignore=DL3008
+RUN \
+--mount=type=cache,id=corepack-cache-${TARGETPLATFORM},target=/usr/local/share/.cache/corepack,sharing=locked \
+--mount=type=cache,id=yarn-cache-${TARGETPLATFORM},target=/usr/local/share/.cache/yarn,sharing=locked \
+# Install Node packages
+  yarn workspaces focus --production @mastodon/mastodon;
 
-RUN yarn workspaces focus --all --production && \
+# Create temporary assets build layer from build layer
-    yarn cache clean
+FROM build as precompiler
 
-FROM node:${NODE_VERSION}
+# Copy Mastodon sources into precompiler layer
+COPY . /opt/mastodon/
 
-# Use those args to specify your own version flags & suffixes
+# Copy bundler and node packages from build layer to container
-ARG MASTODON_VERSION_PRERELEASE=""
+COPY --from=yarn /opt/mastodon /opt/mastodon/
-ARG MASTODON_VERSION_METADATA=""
+COPY --from=bundler /opt/mastodon /opt/mastodon/
+COPY --from=bundler /usr/local/bundle/ /usr/local/bundle/
 
-ARG UID="991"
+ARG TARGETPLATFORM
-ARG GID="991"
 
-COPY --link --from=ruby /opt/ruby /opt/ruby
+RUN \
+# Use Ruby on Rails to create Mastodon assets
+  OTP_SECRET=precompile_placeholder SECRET_KEY_BASE=precompile_placeholder bundle exec rails assets:precompile; \
+# Cleanup temporary files
+  rm -fr /opt/mastodon/tmp;
 
-SHELL ["/bin/bash", "-o", "pipefail", "-c"]
+# Prep final Mastodon Ruby layer
+FROM ruby as mastodon
 
-ENV DEBIAN_FRONTEND="noninteractive" \
+ARG TARGETPLATFORM
-    PATH="${PATH}:/opt/ruby/bin:/opt/mastodon/bin"
 
-# Ignoring these here since we don't want to pin any versions and the Debian image removes apt-get content after use
+# hadolint ignore=DL3008
-# hadolint ignore=DL3008,DL3009
+RUN \
-RUN apt-get update && \
+# Mount Apt cache and lib directories from Docker buildx caches
-    echo "Etc/UTC" > /etc/localtime && \
+--mount=type=cache,id=apt-cache-${TARGETPLATFORM},target=/var/cache/apt,sharing=locked \
-    groupadd -g "${GID}" mastodon && \
+--mount=type=cache,id=apt-lib-${TARGETPLATFORM},target=/var/lib/apt,sharing=locked \
-    useradd -l -u "$UID" -g "${GID}" -m -d /opt/mastodon mastodon && \
+# Mount Corepack and Yarn caches from Docker buildx caches
-    apt-get -y --no-install-recommends install whois \
+--mount=type=cache,id=corepack-cache-${TARGETPLATFORM},target=/usr/local/share/.cache/corepack,sharing=locked \
-        wget \
+--mount=type=cache,id=yarn-cache-${TARGETPLATFORM},target=/usr/local/share/.cache/yarn,sharing=locked \
-        procps \
+# Apt update install non-dev versions of necessary components
-        libssl3 \
+  apt-get update; \
-        libpq5 \
+  apt-get install -y --no-install-recommends \
-        imagemagick \
+    libssl3 \
-        ffmpeg \
+    libpq5 \
-        libjemalloc2 \
+    libicu72 \
-        libicu72 \
+    libidn12 \
-        libidn12 \
+    libreadline8 \
-        libyaml-0-2 \
+    libyaml-0-2 \
-        file \
+  ;
-        ca-certificates \
-        tzdata \
-        libreadline8 \
-        tini && \
-    ln -s /opt/mastodon /mastodon && \
-    corepack enable
 
-# Note: no, cleaning here since Debian does this automatically
+# Copy Mastodon sources into final layer
-# See the file /etc/apt/apt.conf.d/docker-clean within the Docker image's filesystem
+COPY . /opt/mastodon/
 
-COPY --chown=mastodon:mastodon . /opt/mastodon
+# Copy compiled assets to layer
-COPY --chown=mastodon:mastodon --from=build /opt/mastodon /opt/mastodon
+COPY --from=precompiler /opt/mastodon/public/packs /opt/mastodon/public/packs
+COPY --from=precompiler /opt/mastodon/public/assets /opt/mastodon/public/assets
+# Copy bundler components to layer
+COPY --from=bundler /usr/local/bundle/ /usr/local/bundle/
 
-ENV RAILS_ENV="production" \
+RUN \
-    NODE_ENV="production" \
+# Precompile bootsnap code for faster Rails startup
-    RAILS_SERVE_STATIC_FILES="true" \
+  bundle exec bootsnap precompile --gemfile app/ lib/;
-    BIND="0.0.0.0" \
-    MASTODON_VERSION_PRERELEASE="${MASTODON_VERSION_PRERELEASE}" \
-    MASTODON_VERSION_METADATA="${MASTODON_VERSION_METADATA}"
 
-# Set the run user
+RUN \
+# Pre-create and chown system volume to Mastodon user
+  mkdir -p /opt/mastodon/public/system; \
+  chown mastodon:mastodon /opt/mastodon/public/system;
+
+# Set the running user for resulting container
 USER mastodon
-WORKDIR /opt/mastodon
+# Expose default Puma ports
-
+EXPOSE 3000
-# Precompile assets
+# Set container tini as default entry point
-RUN OTP_SECRET=precompile_placeholder SECRET_KEY_BASE=precompile_placeholder rails assets:precompile
+ENTRYPOINT ["/usr/bin/tini", "--"]
-
-# Set the work dir and the container entry point
-ENTRYPOINT ["/usr/bin/tini", "--"]
-EXPOSE 3000 4000

streaming/.dockerignore

@@ -0,0 +1,7 @@
+.env
+.env.*
+.gitignore
+node_modules
+.DS_Store
+*.swp
+*~

streaming/Dockerfile

@@ -0,0 +1,102 @@
+# syntax=docker/dockerfile:1.4
+
+# Please see https://docs.docker.com/engine/reference/builder for information about
+# the extended buildx capabilities used in this file.
+# Make sure multiarch TARGETPLATFORM is available for interpolation
+# See: https://docs.docker.com/build/building/multi-platform/
+ARG TARGETPLATFORM=${TARGETPLATFORM}
+ARG BUILDPLATFORM=${BUILDPLATFORM}
+
+# Node version to use in base image, change with [--build-arg NODE_MAJOR_VERSION="20"]
+ARG NODE_MAJOR_VERSION="20"
+# Debian image to use for base image, change with [--build-arg DEBIAN_VERSION="bookworm"]
+ARG DEBIAN_VERSION="bookworm"
+# Node image to use for base image based on combined variables (ex: 20-bookworm-slim)
+FROM docker.io/node:${NODE_MAJOR_VERSION}-${DEBIAN_VERSION}-slim as streaming
+
+# Timezone used by the Docker container and runtime, change with [--build-arg TZ=Europe/Berlin]
+ARG TZ="Etc/UTC"
+# Linux UID (user id) for the mastodon user, change with [--build-arg UID=1234]
+ARG UID="991"
+# Linux GID (group id) for the mastodon user, change with [--build-arg GID=1234]
+ARG GID="991"
+
+# Apply Mastodon build options based on options above
+ENV \
+# Apply Mastodon version information
+  MASTODON_VERSION_PRERELEASE="${MASTODON_VERSION_PRERELEASE}" \
+  MASTODON_VERSION_METADATA="${MASTODON_VERSION_METADATA}" \
+# Apply timezone
+  TZ=${TZ}
+
+ENV \
+# Configure the IP to bind Mastodon to when serving traffic
+  BIND="0.0.0.0" \
+# Explicitly set PORT to match the exposed port
+  PORT=4000 \
+# Use production settings for Yarn, Node and related nodejs based tools
+  NODE_ENV="production" \
+# Add Ruby and Mastodon installation to the PATH
+  DEBIAN_FRONTEND="noninteractive"
+
+# Set default shell used for running commands
+SHELL ["/bin/bash", "-o", "pipefail", "-o", "errexit", "-c"]
+
+ARG TARGETPLATFORM
+
+RUN echo "Target platform is ${TARGETPLATFORM}"
+
+RUN \
+# Sets timezone
+  echo "${TZ}" > /etc/localtime; \
+# Creates mastodon user/group and sets home directory
+  groupadd -g "${GID}" mastodon; \
+  useradd -l -u "${UID}" -g "${GID}" -m -d /opt/mastodon mastodon; \
+# Creates symlink for /mastodon folder
+  ln -s /opt/mastodon /mastodon;
+
+# hadolint ignore=DL3008,DL3005
+RUN \
+# Mount Apt cache and lib directories from Docker buildx caches
+--mount=type=cache,id=apt-cache-${TARGETPLATFORM},target=/var/cache/apt,sharing=locked \
+--mount=type=cache,id=apt-lib-${TARGETPLATFORM},target=/var/lib/apt,sharing=locked \
+# upgrade to check for security updates to Debian image
+  apt-get update; \
+  apt-get dist-upgrade -yq; \
+  apt-get install -y --no-install-recommends \
+    ca-certificates \
+    curl \
+    tzdata \
+  ;
+
+# Set /opt/mastodon as working directory
+WORKDIR /opt/mastodon
+
+# Copy Node package configuration files from build system to container
+COPY package.json yarn.lock .yarnrc.yml /opt/mastodon/
+COPY .yarn /opt/mastodon/.yarn
+# Copy Streaming source code from build system to container
+COPY ./streaming /opt/mastodon/streaming
+
+RUN \
+# Mount local Corepack and Yarn caches from Docker buildx caches
+--mount=type=cache,id=corepack-cache-${TARGETPLATFORM},target=/usr/local/share/.cache/corepack,sharing=locked \
+--mount=type=cache,id=yarn-cache-${TARGETPLATFORM},target=/usr/local/share/.cache/yarn,sharing=locked \
+  # Configure Corepack
+  rm /usr/local/bin/yarn*; \
+  corepack enable; \
+  corepack prepare --activate;
+
+RUN \
+# Mount Corepack and Yarn caches from Docker buildx caches
+--mount=type=cache,id=corepack-cache-${TARGETPLATFORM},target=/usr/local/share/.cache/corepack,sharing=locked \
+--mount=type=cache,id=yarn-cache-${TARGETPLATFORM},target=/usr/local/share/.cache/yarn,sharing=locked \
+# Install Node packages
+  yarn workspaces focus --production @mastodon/streaming;
+
+# Set the running user for resulting container
+USER mastodon
+# Expose default Streaming ports
+EXPOSE 4000
+# Run streaming when started
+CMD [ node ./streaming/index.js ]