Files
eagle0/.github/workflows/docker_build.yml
T
ddbc960945 Add aggressive docker image cleanup to prevent disk full (#5841)
The existing cleanup only removed dangling images. Now also removes
images older than 24h to prevent disk space exhaustion while keeping
recent images available for rollback.

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2026-02-04 09:10:35 -08:00

410 lines
18 KiB
YAML

name: Docker Build and Push
on:
push:
branches: [ "main" ]
paths:
# Note: C++ changes trigger shardok_arm64_build.yml instead
# Note: Auth changes trigger auth_build.yml instead
# Note: Windows installer changes trigger installer_build.yml instead
- 'src/main/go/**'
- '!src/main/go/net/eagle0/authservice/**'
- '!src/main/go/net/eagle0/authcli/**'
- '!src/main/go/net/eagle0/clients/**'
- 'src/main/scala/**'
- 'src/main/protobuf/**'
- '!src/main/protobuf/net/eagle0/eagle/api/auth.proto'
- '!src/main/protobuf/net/eagle0/eagle/api/admin/**'
- '!src/main/protobuf/net/eagle0/eagle/internal/auth_internal.proto'
- 'src/main/resources/**'
- 'ci/BUILD.bazel'
- 'MODULE.bazel'
- 'docker-compose.prod.yml'
- 'nginx/**'
- '.github/workflows/docker_build.yml'
workflow_dispatch:
inputs:
push_images:
description: 'Push images to container registry'
required: true
default: 'false'
type: boolean
# Only allow one deployment at a time to prevent race conditions
concurrency:
group: docker-build-deploy
cancel-in-progress: false # Don't cancel running deployments, queue new ones
permissions:
contents: read
jobs:
# Single consolidated build job - builds all images with one bazel invocation
# This uses 1 runner slot instead of 4, and Bazel parallelizes internally
# NOTE: Must run on a runner with the bazel label to ensure consistent Bazel cache
build-all:
runs-on: [self-hosted, bazel]
outputs:
eagle_image_tag: ${{ steps.push-images.outputs.eagle_image_tag }}
admin_image_tag: ${{ steps.push-images.outputs.admin_image_tag }}
jfr_sidecar_image_tag: ${{ steps.push-images.outputs.jfr_sidecar_image_tag }}
steps:
- name: Checkout repository
uses: actions/checkout@v4
with:
lfs: false
- name: Build all Docker images
id: build-all
run: |
set -ex
# Build ALL images in a single bazel command - Bazel parallelizes internally
# Note: Shardok is built separately for ARM64 and deployed to Hetzner
echo "=== Building Docker images ==="
bazel build \
--platforms=//:linux_x86_64 \
--extra_toolchains=@llvm_toolchain_linux//:all \
//ci:eagle_server_image \
//ci:admin_server_image \
//ci:jfr_sidecar_image \
//ci:warmup_tar
# Extract warmup binary from tar for deployment
mkdir -p scripts/bin
tar -xf bazel-bin/ci/warmup_tar.tar -C scripts/bin --strip-components=1
# Save all image paths before any other bazel command changes bazel-bin symlink
EAGLE_PATH=$(readlink -f bazel-bin/ci/eagle_server_image)
ADMIN_PATH=$(readlink -f bazel-bin/ci/admin_server_image)
JFR_PATH=$(readlink -f bazel-bin/ci/jfr_sidecar_image)
echo "eagle_path=$EAGLE_PATH" >> $GITHUB_OUTPUT
echo "admin_path=$ADMIN_PATH" >> $GITHUB_OUTPUT
echo "jfr_path=$JFR_PATH" >> $GITHUB_OUTPUT
echo "=== Image paths ==="
echo "Eagle: $EAGLE_PATH"
echo "Admin: $ADMIN_PATH"
echo "JFR Sidecar: $JFR_PATH"
- name: Login to DigitalOcean Container Registry
if: github.event_name == 'push' || (github.event_name == 'workflow_dispatch' && github.event.inputs.push_images == 'true')
env:
DO_TOKEN: ${{ secrets.DO_REGISTRY_TOKEN }}
run: |
mkdir -p ~/.docker
AUTH=$(echo -n "${DO_TOKEN}:${DO_TOKEN}" | base64)
echo "{\"auths\":{\"registry.digitalocean.com\":{\"auth\":\"${AUTH}\"}}}" > ~/.docker/config.json
mkdir -p .docker
cp ~/.docker/config.json .docker/
- name: Push all images to DO registry
id: push-images
if: github.event_name == 'push' || (github.event_name == 'workflow_dispatch' && github.event.inputs.push_images == 'true')
env:
DOCKER_CONFIG: ${{ github.workspace }}/.docker
run: |
set -ex
GIT_SHA=$(git rev-parse --short=8 HEAD)
# Get crane from push target runfiles
bazel build //ci:eagle_server_push
CRANE="bazel-bin/ci/push_eagle_server_push.sh.runfiles/rules_oci~~oci~oci_crane_darwin_arm64/crane"
if [ ! -e "$CRANE" ]; then
# Fallback: find any Darwin crane
RUNFILES="bazel-bin/ci/push_eagle_server_push.sh.runfiles"
CRANE=$(find "$RUNFILES" -path "*darwin*" -name crane 2>/dev/null | head -1)
fi
if [ -z "$CRANE" ] || [ ! -e "$CRANE" ]; then
echo "ERROR: crane not found"
exit 1
fi
echo "Using crane: $CRANE"
# Push Eagle image
EAGLE_IMAGE="${{ steps.build-all.outputs.eagle_path }}"
EAGLE_TAG="registry.digitalocean.com/eagle0/eagle-server:${GIT_SHA}"
echo "Pushing Eagle: $EAGLE_TAG"
$CRANE push "$EAGLE_IMAGE" "$EAGLE_TAG"
$CRANE copy "$EAGLE_TAG" "registry.digitalocean.com/eagle0/eagle-server:latest"
echo "eagle_image_tag=$EAGLE_TAG" >> $GITHUB_OUTPUT
# Push Admin image
ADMIN_IMAGE="${{ steps.build-all.outputs.admin_path }}"
ADMIN_TAG="registry.digitalocean.com/eagle0/admin-server:${GIT_SHA}"
echo "Pushing Admin: $ADMIN_TAG"
$CRANE push "$ADMIN_IMAGE" "$ADMIN_TAG"
$CRANE copy "$ADMIN_TAG" "registry.digitalocean.com/eagle0/admin-server:latest"
echo "admin_image_tag=$ADMIN_TAG" >> $GITHUB_OUTPUT
# Push JFR Sidecar image
JFR_IMAGE="${{ steps.build-all.outputs.jfr_path }}"
JFR_TAG="registry.digitalocean.com/eagle0/jfr-sidecar:${GIT_SHA}"
echo "Pushing JFR Sidecar: $JFR_TAG"
$CRANE push "$JFR_IMAGE" "$JFR_TAG"
$CRANE copy "$JFR_TAG" "registry.digitalocean.com/eagle0/jfr-sidecar:latest"
echo "jfr_sidecar_image_tag=$JFR_TAG" >> $GITHUB_OUTPUT
echo "=== All images pushed successfully ==="
deploy:
runs-on: [self-hosted, bazel]
needs: [build-all]
if: github.event_name == 'push' || (github.event_name == 'workflow_dispatch' && github.event.inputs.push_images == 'true')
environment: production
env:
EAGLE_IMAGE: ${{ needs.build-all.outputs.eagle_image_tag }}
ADMIN_IMAGE: ${{ needs.build-all.outputs.admin_image_tag }}
JFR_SIDECAR_IMAGE: ${{ needs.build-all.outputs.jfr_sidecar_image_tag }}
OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
GEMINI_API_KEY: ${{ secrets.GEMINI_API_KEY }}
GPT_MODEL_NAME: ${{ secrets.GPT_MODEL_NAME }}
EAGLE_ENABLE_S3: ${{ secrets.EAGLE_ENABLE_S3 }}
DO_SPACES_ACCESS_KEY: ${{ secrets.DO_SPACES_ACCESS_KEY }}
DO_SPACES_SECRET_KEY: ${{ secrets.DO_SPACES_SECRET_KEY }}
JWT_PRIVATE_KEY: ${{ secrets.JWT_PRIVATE_KEY }}
DISCORD_CLIENT_ID: ${{ secrets.DISCORD_CLIENT_ID }}
DISCORD_CLIENT_SECRET: ${{ secrets.DISCORD_CLIENT_SECRET }}
GOOGLE_CLIENT_ID: ${{ secrets.GOOGLE_CLIENT_ID }}
GOOGLE_CLIENT_SECRET: ${{ secrets.GOOGLE_CLIENT_SECRET }}
GH_OAUTH_CLIENT_ID: ${{ secrets.GH_OAUTH_CLIENT_ID }}
GH_OAUTH_CLIENT_SECRET: ${{ secrets.GH_OAUTH_CLIENT_SECRET }}
APPLE_SIGNIN_CLIENT_ID: ${{ secrets.APPLE_SIGNIN_CLIENT_ID }}
APPLE_TEAM_ID: ${{ secrets.APPLE_TEAM_ID }}
APPLE_SIGNIN_KEY_ID: ${{ secrets.APPLE_SIGNIN_KEY_ID }}
APPLE_SIGNIN_PRIVATE_KEY: ${{ secrets.APPLE_SIGNIN_PRIVATE_KEY }}
TWITCH_CLIENT_ID: ${{ secrets.TWITCH_CLIENT_ID }}
TWITCH_CLIENT_SECRET: ${{ secrets.TWITCH_CLIENT_SECRET }}
SHARDOK_ADDRESS: ${{ secrets.SHARDOK_ADDRESS }}
SHARDOK_AUTH_TOKEN: ${{ secrets.SHARDOK_AUTH_TOKEN }}
SENTRY_DSN: ${{ secrets.SENTRY_DSN }}
FASTMAIL_API_TOKEN: ${{ secrets.FASTMAIL_API_TOKEN }}
FASTMAIL_FROM_EMAIL: ${{ secrets.FASTMAIL_FROM_EMAIL }}
FASTMAIL_FROM_NAME: ${{ secrets.FASTMAIL_FROM_NAME }}
DO_DROPLET_IP: ${{ secrets.DO_DROPLET_IP }}
DO_REGISTRY_TOKEN: ${{ secrets.DO_REGISTRY_TOKEN }}
NOTIFY_SECRET: ${{ secrets.EAGLE_NOTIFY_SECRET }}
steps:
- name: Checkout repository
uses: actions/checkout@v4
- name: Setup SSH key
run: |
mkdir -p ~/.ssh
echo "${{ secrets.DO_SSH_KEY }}" > ~/.ssh/deploy_key
chmod 600 ~/.ssh/deploy_key
ssh-keyscan -H "$DO_DROPLET_IP" >> ~/.ssh/known_hosts 2>/dev/null || true
- name: Build warmup tool
run: |
bazel build //ci:warmup_tar
mkdir -p scripts/bin
tar -xf bazel-bin/ci/warmup_tar.tar -C scripts/bin --strip-components=1
- name: Copy config files to droplet
run: |
# Create directory structure on remote
ssh -i ~/.ssh/deploy_key deploy@"$DO_DROPLET_IP" bash -s << 'SETUP_DIRS'
set -e
mkdir -p /opt/eagle0/scripts/bin /opt/eagle0/nginx
rm -f /opt/eagle0/scripts/bin/warmup
SETUP_DIRS
# Copy files
scp -i ~/.ssh/deploy_key docker-compose.prod.yml deploy@"$DO_DROPLET_IP":/opt/eagle0/
scp -i ~/.ssh/deploy_key nginx/nginx.conf deploy@"$DO_DROPLET_IP":/opt/eagle0/nginx/
scp -i ~/.ssh/deploy_key scripts/deploy-blue-green.sh scripts/warmup-eagle.sh scripts/eagle-exec.sh scripts/eagle-logs.sh deploy@"$DO_DROPLET_IP":/opt/eagle0/scripts/
scp -i ~/.ssh/deploy_key scripts/bin/warmup deploy@"$DO_DROPLET_IP":/opt/eagle0/scripts/bin/
- name: Deploy to production droplet
run: |
ssh -i ~/.ssh/deploy_key deploy@"$DO_DROPLET_IP" bash -s << DEPLOY_SCRIPT
set -ex
cd /opt/eagle0
# =================================================================
# CRITICAL: Validate environment variables before proceeding
# This catches GitHub Actions secret store hiccups early
# =================================================================
validate_env() {
local var_name="\$1"
local var_value="\$2"
local default_value="\${3:-}"
if [ -z "\${var_value}" ]; then
echo "ERROR: \${var_name} is empty. GitHub Actions secrets may have failed to load."
echo "Please retry the workflow."
return 1
fi
# Check if we got a default value instead of the real secret
if [ -n "\${default_value}" ] && [ "\${var_value}" = "\${default_value}" ]; then
echo "ERROR: \${var_name} has default value '\${default_value}' instead of the actual secret."
echo "This indicates GitHub Actions secrets failed to load. Please retry the workflow."
return 1
fi
return 0
}
echo "Validating critical environment variables..."
# These are the raw values from GitHub Actions (before export)
# We check them before exporting to catch issues early
VALIDATION_FAILED=0
validate_env "SHARDOK_ADDRESS" "${SHARDOK_ADDRESS}" "" || VALIDATION_FAILED=1
validate_env "EAGLE_IMAGE" "${EAGLE_IMAGE}" "" || VALIDATION_FAILED=1
validate_env "JWT_PRIVATE_KEY" "${JWT_PRIVATE_KEY}" "" || VALIDATION_FAILED=1
validate_env "DO_REGISTRY_TOKEN" "${DO_REGISTRY_TOKEN}" "" || VALIDATION_FAILED=1
if [ "\${VALIDATION_FAILED}" -eq 1 ]; then
echo ""
echo "========================================="
echo "DEPLOYMENT ABORTED: Missing critical secrets"
echo "This is likely a transient GitHub Actions issue."
echo "Please retry the workflow."
echo "========================================="
exit 1
fi
echo "All critical environment variables validated successfully."
# =================================================================
# Export environment variables for docker compose
# These are passed via heredoc and exported so child processes (docker compose) can access them
export EAGLE_IMAGE="${EAGLE_IMAGE}"
export ADMIN_IMAGE="${ADMIN_IMAGE}"
export JFR_SIDECAR_IMAGE="${JFR_SIDECAR_IMAGE}"
export OPENAI_API_KEY="${OPENAI_API_KEY}"
export ANTHROPIC_API_KEY="${ANTHROPIC_API_KEY}"
export GEMINI_API_KEY="${GEMINI_API_KEY}"
export GPT_MODEL_NAME="${GPT_MODEL_NAME:-gpt-4o}"
export EAGLE_ENABLE_S3="${EAGLE_ENABLE_S3:-false}"
export DO_SPACES_ACCESS_KEY="${DO_SPACES_ACCESS_KEY}"
export DO_SPACES_SECRET_KEY="${DO_SPACES_SECRET_KEY}"
export JWT_PRIVATE_KEY="${JWT_PRIVATE_KEY}"
export DISCORD_CLIENT_ID="${DISCORD_CLIENT_ID}"
export DISCORD_CLIENT_SECRET="${DISCORD_CLIENT_SECRET}"
export GOOGLE_CLIENT_ID="${GOOGLE_CLIENT_ID}"
export GOOGLE_CLIENT_SECRET="${GOOGLE_CLIENT_SECRET}"
export GH_OAUTH_CLIENT_ID="${GH_OAUTH_CLIENT_ID}"
export GH_OAUTH_CLIENT_SECRET="${GH_OAUTH_CLIENT_SECRET}"
export APPLE_SIGNIN_CLIENT_ID="${APPLE_SIGNIN_CLIENT_ID}"
export APPLE_TEAM_ID="${APPLE_TEAM_ID}"
export APPLE_SIGNIN_KEY_ID="${APPLE_SIGNIN_KEY_ID}"
export APPLE_SIGNIN_PRIVATE_KEY="${APPLE_SIGNIN_PRIVATE_KEY}"
export TWITCH_CLIENT_ID="${TWITCH_CLIENT_ID}"
export TWITCH_CLIENT_SECRET="${TWITCH_CLIENT_SECRET}"
export SHARDOK_ADDRESS="${SHARDOK_ADDRESS}"
export SHARDOK_AUTH_TOKEN="${SHARDOK_AUTH_TOKEN}"
export SENTRY_DSN="${SENTRY_DSN}"
export FASTMAIL_API_TOKEN="${FASTMAIL_API_TOKEN}"
export FASTMAIL_FROM_EMAIL="${FASTMAIL_FROM_EMAIL}"
export FASTMAIL_FROM_NAME="${FASTMAIL_FROM_NAME}"
export DO_REGISTRY_TOKEN="${DO_REGISTRY_TOKEN}"
export NOTIFY_SECRET="${NOTIFY_SECRET}"
# Check Docker has IPv6 support
if ! cat /etc/docker/daemon.json 2>/dev/null | grep -q '"ip6tables"'; then
echo "WARNING: Docker IPv6 not configured. Eagle may not reach Hetzner Shardok."
fi
# Login to registry
echo "\$DO_REGISTRY_TOKEN" | docker login registry.digitalocean.com -u "\$DO_REGISTRY_TOKEN" --password-stdin
echo "Using images: \$EAGLE_IMAGE, \$ADMIN_IMAGE, \$JFR_SIDECAR_IMAGE"
# Install crane for pulling OCI images
echo "Installing crane..."
rm -f crane
curl -sL https://github.com/google/go-containerregistry/releases/download/v0.20.2/go-containerregistry_Linux_x86_64.tar.gz | tar xzf - crane
chmod +x crane
# Pull and load all images
echo "Pulling Eagle image..."
./crane pull "\${EAGLE_IMAGE}" eagle.tar && docker load -i eagle.tar && rm eagle.tar
echo "Pulling Admin image..."
./crane pull "\${ADMIN_IMAGE}" admin.tar && docker load -i admin.tar && rm admin.tar
# Tag as :latest locally so docker-compose fallback uses correct image
docker tag "\${ADMIN_IMAGE}" registry.digitalocean.com/eagle0/admin-server:latest
echo "Pulling JFR Sidecar image..."
./crane pull "\${JFR_SIDECAR_IMAGE}" jfr-sidecar.tar && docker load -i jfr-sidecar.tar && rm jfr-sidecar.tar
# Pull other compose images
docker pull nginx:alpine || true
docker pull certbot/certbot || true
echo "All images pulled successfully"
# =================================================================
# Verify Shardok connectivity before proceeding with deployment
# This catches network/firewall issues early
# =================================================================
echo "Verifying Shardok connectivity..."
SHARDOK_HOST=\$(echo "\${SHARDOK_ADDRESS}" | cut -d: -f1)
SHARDOK_PORT=\$(echo "\${SHARDOK_ADDRESS}" | cut -d: -f2)
# Try to connect to Shardok (timeout after 10 seconds)
if nc -z -w 10 "\${SHARDOK_HOST}" "\${SHARDOK_PORT}" 2>/dev/null; then
echo "Shardok connectivity verified: \${SHARDOK_ADDRESS} is reachable"
else
echo ""
echo "========================================="
echo "ERROR: Cannot reach Shardok at \${SHARDOK_ADDRESS}"
echo "This may indicate:"
echo " - Shardok server is not running on Hetzner"
echo " - Network/firewall issues between DigitalOcean and Hetzner"
echo " - Incorrect SHARDOK_ADDRESS configuration"
echo ""
echo "DEPLOYMENT ABORTED: Shardok must be reachable for battles to work."
echo "========================================="
exit 1
fi
# Stop local shardok container if running (now runs on Hetzner)
docker stop shardok-server 2>/dev/null || true
docker rm shardok-server 2>/dev/null || true
# Deploy Eagle with blue-green (handles eagle, nginx, admin, jfr-sidecar)
# Note: Shardok runs on Hetzner, deployed separately via shardok_arm64_build.yml
# Note: Auth is deployed separately via auth_build.yml - do NOT touch auth here
chmod +x /opt/eagle0/scripts/*.sh
[ -f "/opt/eagle0/scripts/bin/warmup" ] && chmod +x /opt/eagle0/scripts/bin/warmup
GIT_SHA=\$(echo "\${EAGLE_IMAGE}" | sed 's/.*://')
/opt/eagle0/scripts/deploy-blue-green.sh "\${GIT_SHA}"
# Verify
sleep 10
docker compose -f docker-compose.prod.yml ps
docker compose -f docker-compose.prod.yml images
# Verify admin container is running the correct image
echo "=== Verifying admin container image ==="
ADMIN_RUNNING_DIGEST=\$(docker inspect admin-server --format '{{.Image}}')
ADMIN_EXPECTED_DIGEST=\$(docker inspect "\${ADMIN_IMAGE}" --format '{{.Id}}')
echo "Expected image: \${ADMIN_IMAGE}"
echo "Expected digest: \${ADMIN_EXPECTED_DIGEST}"
echo "Running digest: \${ADMIN_RUNNING_DIGEST}"
if [ "\${ADMIN_RUNNING_DIGEST}" != "\${ADMIN_EXPECTED_DIGEST}" ]; then
echo "ERROR: Admin container is running wrong image!"
echo "Container entrypoint:"
docker inspect admin-server --format '{{.Config.Entrypoint}}'
exit 1
fi
echo "Admin image verification passed"
# Cleanup - remove old images to prevent disk full
docker container prune -f
docker image prune -f
# Remove images older than 24h (keeps recent images for rollback)
docker image prune -a -f --filter "until=24h" || true
DEPLOY_SCRIPT