mirror of
https://github.com/jhd3197/ServerKit.git
synced 2026-10-23 03:27:43 +09:00
Juan Denis
75bd900442
The Dockerfile allowed 5s before failed health probes count, while compose allows 40s; boot runs migrations and settings seeding, so a plain docker run could mark a healthy first boot unhealthy.
194 lines
8.4 KiB
Docker
194 lines
8.4 KiB
Docker
# ============================================
|
|
# ServerKit Dockerfile - Single Container
|
|
# ============================================
|
|
# This Dockerfile creates a single container with both
|
|
# frontend and backend. Good for simple deployments.
|
|
#
|
|
# For production with separate containers and nginx,
|
|
# use docker-compose.yml instead.
|
|
#
|
|
# Build: docker build -t serverkit .
|
|
# Run: docker run -d -p 5000:5000 --env-file .env serverkit
|
|
# Override the internal backend port with SERVERKIT_BACKEND_PORT if needed.
|
|
#
|
|
# Build args:
|
|
# INSTALL_CLAMAV=true bundle the ClamAV scanner (~250 MB). Off by default:
|
|
# malware scanning falls back to the builtin YARA/pure-
|
|
# Python matcher and reports clamav_available=false, so
|
|
# the feature degrades instead of breaking. Operators who
|
|
# want ClamAV can build with it or apt-install it beside
|
|
# the panel.
|
|
# ============================================
|
|
|
|
# Stage 1: Build Frontend
|
|
# --platform=$BUILDPLATFORM: dist/ is static JS/CSS (arch-independent), so this
|
|
# stage always runs natively on the build machine — never under QEMU, where an
|
|
# emulated `npm ci` takes 20-40 min (or hangs) during multi-arch release builds.
|
|
FROM --platform=$BUILDPLATFORM node:22-alpine AS frontend-builder
|
|
|
|
WORKDIR /app/frontend
|
|
|
|
# Copy package files
|
|
COPY frontend/package*.json ./
|
|
|
|
# Install dependencies
|
|
RUN npm ci
|
|
|
|
# Copy frontend source
|
|
COPY frontend/ ./
|
|
|
|
# Build production bundle
|
|
RUN npm run build
|
|
|
|
# ============================================
|
|
# Stage 2: Python dependency builder
|
|
#
|
|
# The compiler toolchain (gcc, *-dev headers) is needed to build wheels for
|
|
# cryptography / lxml / psutil on platforms without a prebuilt wheel, but it is
|
|
# ~280 MB and has no business in a running panel. Building into a self-contained
|
|
# venv here lets the runtime stage take the result and leave the toolchain behind.
|
|
FROM python:3.11-slim-bookworm AS python-builder
|
|
|
|
ENV PYTHONDONTWRITEBYTECODE=1 \
|
|
PYTHONUNBUFFERED=1
|
|
|
|
RUN apt-get update && apt-get install -y --no-install-recommends \
|
|
gcc \
|
|
python3-dev \
|
|
# Required for cryptography
|
|
libffi-dev \
|
|
libssl-dev \
|
|
&& rm -rf /var/lib/apt/lists/*
|
|
|
|
RUN python -m venv /opt/venv
|
|
ENV PATH="/opt/venv/bin:$PATH"
|
|
|
|
COPY backend/requirements.txt /tmp/requirements.txt
|
|
RUN pip install --no-cache-dir --upgrade pip \
|
|
&& pip install --no-cache-dir -r /tmp/requirements.txt
|
|
|
|
# Trim the venv (~50 MB). Both prunes happen here in the builder, so a mistake
|
|
# surfaces as a failed image build, never as a half-installed panel on a host.
|
|
#
|
|
# 1. Bundled test suites. Third-party packages ship their own pytest trees
|
|
# (passlib is the big one) that no running panel imports.
|
|
# 2. botocore API models. botocore carries JSON models for 400+ AWS services
|
|
# (~15 MB); ServerKit constructs exactly three clients — s3
|
|
# (storage_provider_service), route53 (dns_provider_service) and ses
|
|
# (notifications/providers) — plus sts, which botocore resolves while signing.
|
|
# The glob below matches directories only, so botocore's top-level
|
|
# endpoints.json and partitions.json are never touched.
|
|
#
|
|
# >>> Adding a new boto3.client('<svc>') call? Add <svc> to BOTOCORE_KEEP or
|
|
# >>> it raises UnknownServiceError at runtime — in the image only, never in
|
|
# >>> dev. backend/tests/test_botocore_prune_contract.py parses this ARG and
|
|
# >>> every call site and fails the suite when the two drift apart.
|
|
ARG BOTOCORE_KEEP="s3 route53 ses sts"
|
|
RUN set -eux; \
|
|
SP=/opt/venv/lib/python3.11/site-packages; \
|
|
find "$SP" -type d \( -name tests -o -name test \) -prune -exec rm -rf {} +; \
|
|
for d in "$SP"/botocore/data/*/; do \
|
|
name="$(basename "$d")"; \
|
|
case " $BOTOCORE_KEEP " in *" $name "*) ;; *) rm -rf "$d";; esac; \
|
|
done
|
|
|
|
# ============================================
|
|
# Stage 3: Production Image
|
|
FROM python:3.11-slim-bookworm
|
|
|
|
ARG INSTALL_CLAMAV=false
|
|
|
|
# Set environment variables
|
|
# PATH puts the venv first so `python`, `pip` and `gunicorn` all resolve to it —
|
|
# including the runtime `sys.executable -m pip install` that PluginService uses
|
|
# to install an extension's Python requirements.
|
|
ENV PYTHONDONTWRITEBYTECODE=1 \
|
|
PYTHONUNBUFFERED=1 \
|
|
FLASK_ENV=production \
|
|
SERVERKIT_BACKEND_PORT=5000 \
|
|
PATH="/opt/venv/bin:$PATH"
|
|
|
|
# Runtime system dependencies only — no compiler, no -dev headers. psutil and
|
|
# cryptography link against libssl3/libffi8, which the slim base already ships.
|
|
RUN apt-get update && apt-get install -y --no-install-recommends \
|
|
# Health check
|
|
curl \
|
|
# Useful utilities (ps/top for operators exec'ing into the container)
|
|
procps \
|
|
&& if [ "$INSTALL_CLAMAV" = "true" ]; then \
|
|
apt-get install -y --no-install-recommends \
|
|
clamav clamav-daemon clamav-freshclam; \
|
|
fi \
|
|
&& rm -rf /var/lib/apt/lists/* \
|
|
&& apt-get clean
|
|
|
|
# Create non-root user for security. Declared BEFORE the COPYs below so they can
|
|
# use --chown: a trailing `chown -R` would otherwise rewrite every file's metadata
|
|
# and duplicate the whole tree (~200 MB of venv + app) into an extra layer.
|
|
RUN groupadd -r serverkit && useradd -r -g serverkit serverkit
|
|
|
|
# Bring in the fully-built virtualenv from the builder stage. Owned by serverkit
|
|
# so runtime extension installs (PluginService's `pip install -r`) can write to it.
|
|
COPY --from=python-builder --chown=serverkit:serverkit /opt/venv /opt/venv
|
|
|
|
# Create necessary directories.
|
|
#
|
|
# /app/instance is where config.py's default DATABASE_URL points
|
|
# (sqlite:////app/instance/serverkit.db). It must exist AND be owned by the
|
|
# serverkit user before the USER switch below: SQLAlchemy will not create the
|
|
# parent directory, so without this the container dies on boot with
|
|
# "sqlite3.OperationalError: unable to open database file". Creating it here
|
|
# also fixes the ownership of a named volume mounted at that path — Docker
|
|
# seeds an empty volume from the image directory, permissions included, but
|
|
# creates it root-owned when the path is absent from the image.
|
|
RUN mkdir -p /etc/serverkit /var/serverkit/apps /var/log/serverkit /var/quarantine /var/backups/serverkit /app/data /app/instance \
|
|
&& chown -R serverkit:serverkit /etc/serverkit /var/serverkit /var/log/serverkit /var/quarantine /var/backups/serverkit /app
|
|
|
|
# Set working directory
|
|
WORKDIR /app
|
|
|
|
# Copy backend source
|
|
COPY --chown=serverkit:serverkit backend/ ./backend/
|
|
|
|
# Ship the VERSION file next to the backend tree (/app/VERSION) so the panel
|
|
# reports its real version in containers instead of the unknown-version fallback
|
|
COPY --chown=serverkit:serverkit VERSION ./VERSION
|
|
|
|
# Ship the fleet-agent installers. The panel serves these over HTTP at
|
|
# /api/v1/servers/install.sh and /api/v1/servers/install.ps1, reading them off
|
|
# its own disk at <tree root>/scripts — which is /app/scripts here. Neither is
|
|
# served at the domain root; spelling the prefix out twice is deliberate, since
|
|
# a half-written path is what sent the #101 reporter hunting for a wrong URL.
|
|
#
|
|
# They were missing from the image entirely, so every containerised panel
|
|
# answered the enrollment one-liner its own UI prints with a 404, and no server
|
|
# could ever join the fleet (#101).
|
|
#
|
|
# Only these two files: both are standalone (they source nothing from
|
|
# scripts/lib), so the rest of scripts/ — lib, keys, the test harnesses,
|
|
# stage-remote.sh — stays out of the image where it belongs.
|
|
COPY --chown=serverkit:serverkit scripts/install.sh scripts/install.ps1 ./scripts/
|
|
|
|
# Copy built frontend from Stage 1
|
|
COPY --from=frontend-builder --chown=serverkit:serverkit /app/frontend/dist ./frontend/dist
|
|
|
|
# Switch to non-root user
|
|
USER serverkit
|
|
|
|
# Expose port
|
|
EXPOSE 5000
|
|
|
|
# Health check. Boot runs schema sync, migrations and settings seeding; the
|
|
# 40s grace matches docker-compose.yml so a slow first boot is not unhealthy.
|
|
HEALTHCHECK --interval=30s --timeout=10s --start-period=40s --retries=3 \
|
|
CMD curl -f "http://localhost:${SERVERKIT_BACKEND_PORT:-5000}/api/v1/system/health" || exit 1
|
|
|
|
# Working directory for the backend
|
|
WORKDIR /app/backend
|
|
|
|
# Default command - use gunicorn for production
|
|
# Single process (agent gateway state is in-memory) + threads: the app runs
|
|
# async_mode='threading' (simple-websocket serves WS); a gevent-websocket
|
|
# worker would double-answer the upgrade handshake and break WebSocket.
|
|
CMD ["sh", "-c", "exec gunicorn --workers 1 --threads 100 --bind 0.0.0.0:${SERVERKIT_BACKEND_PORT:-5000} --timeout 120 --access-logfile - --error-logfile - run:app"]
|