refactor(infra): converge on setup+bootstrap pattern, fix systemd copy bug

- setup_server.sh: full rewrite to match materia/template pattern — adds Docker install, git/curl/ca-certificates apt install, age + sops install (arch-aware), uv install as service user, age keypair generation, SSH config write (root+chown); removes systemd unit copy (was buggy: copied before repo was cloned) - NEW bootstrap_supervisor.sh: ~45 lines — age key check, clone/fetch, tag checkout, sops decrypt, uv sync, copy landing-backup + supervisor systemd units, enable + start - deploy.sh: replace 53-line self-install preamble (sops/age install + keypair generation + exit-1 flow) with simple sops check + decrypt; Docker blue/green logic unchanged Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-02-27 06:57:00 +01:00
parent 8b74d6a76b
commit 5218717e8d
3 changed files with 224 additions and 120 deletions
--- a/deploy.sh
+++ b/deploy.sh
@@ -1,60 +1,17 @@
 #!/usr/bin/env bash
 set -euo pipefail

-# ── Ensure sops + age are installed ───────────────────────
 APP_DIR="$(cd "$(dirname "$0")" && pwd)"
-BIN_DIR="$APP_DIR/bin"
-mkdir -p "$BIN_DIR"
-export PATH="$BIN_DIR:$PATH"

-ARCH=$(uname -m)
-case "$ARCH" in
-    x86_64)  ARCH_SOPS="amd64"; ARCH_AGE="amd64" ;;
-    aarch64) ARCH_SOPS="arm64"; ARCH_AGE="arm64" ;;
-    *) echo "Unsupported architecture: $ARCH"; exit 1 ;;
-esac
-
-if ! command -v age &>/dev/null; then
-    echo "==> Installing age to $BIN_DIR..."
-    AGE_VERSION="v1.3.1"
-    curl -fsSL "https://dl.filippo.io/age/${AGE_VERSION}?for=linux/${ARCH_AGE}" -o /tmp/age.tar.gz
-    tar -xzf /tmp/age.tar.gz -C "$BIN_DIR" --strip-components=1 age/age age/age-keygen
-    chmod +x "$BIN_DIR/age" "$BIN_DIR/age-keygen"
-    rm /tmp/age.tar.gz
-fi
+# ── Decrypt secrets ───────────────────────────────────────

 if ! command -v sops &>/dev/null; then
-    echo "==> Installing sops to $BIN_DIR..."
-    SOPS_VERSION="v3.12.1"
-    curl -fsSL "https://github.com/getsops/sops/releases/download/${SOPS_VERSION}/sops-${SOPS_VERSION}.linux.${ARCH_SOPS}" -o "$BIN_DIR/sops"
-    chmod +x "$BIN_DIR/sops"
-fi
-
-# ── Ensure age keypair exists ─────────────────────────────
-AGE_KEY_FILE="${SOPS_AGE_KEY_FILE:-$APP_DIR/age-key.txt}"
-export SOPS_AGE_KEY_FILE="$AGE_KEY_FILE"
-
-if [ ! -f "$AGE_KEY_FILE" ]; then
-    echo "==> Generating age keypair..."
-    age-keygen -o "$AGE_KEY_FILE" 2>&1
-    chmod 600 "$AGE_KEY_FILE"
-    AGE_PUB=$(grep "public key:" "$AGE_KEY_FILE" | awk '{print $NF}')
-    echo ""
-    echo "!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!"
-    echo "!! NEW SERVER — add this public key to .sops.yaml:    !!"
-    echo "!!                                                     !!"
-    echo "!! $AGE_PUB !!"
-    echo "!!                                                     !!"
-    echo "!! Then run: sops updatekeys .env.prod.sops            !!"
-    echo "!! Commit, push, and re-deploy.                        !!"
-    echo "!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!"
-    echo ""
+    echo "ERROR: sops not found — run infra/setup_server.sh first"
    exit 1
 fi

-# ── Decrypt secrets ───────────────────────────────────────
-sops --input-type dotenv --output-type dotenv -d .env.prod.sops > .env
-chmod 600 .env
+sops --input-type dotenv --output-type dotenv -d "${APP_DIR}/.env.prod.sops" > "${APP_DIR}/.env"
+chmod 600 "${APP_DIR}/.env"

 COMPOSE="docker compose -f docker-compose.prod.yml"
 LIVE_FILE=".live-slot"
--- a/infra/bootstrap_supervisor.sh
+++ b/infra/bootstrap_supervisor.sh
@@ -0,0 +1,72 @@
+#!/bin/bash
+# Bootstrap Padelnomics supervisor after setup_server.sh + adding keys.
+# Run once on the server after SSH deploy key is added to GitLab
+# and the server age key is committed to .env.prod.sops.
+#
+# Usage:
+#   ssh root@<server_ip> 'bash -s' < infra/bootstrap_supervisor.sh
+#
+# Prerequisites:
+#   - setup_server.sh already run (padelnomics_service user, SSH deploy key, age keypair, uv)
+#   - Deploy key added to GitLab (Settings → Repository → Deploy Keys)
+#   - Server age public key added to .sops.yaml + .env.prod.sops committed + pushed
+
+set -euo pipefail
+
+SERVICE_USER="padelnomics_service"
+REPO_DIR="/opt/padelnomics"
+GITLAB_PROJECT="deemanone/padelnomics"
+UV="/home/${SERVICE_USER}/.local/bin/uv"
+
+[ "$(id -u)" = "0" ] || { echo "ERROR: Run as root"; exit 1; }
+
+# ── Check age keypair ─────────────────────────────────────────────────────────
+
+AGE_KEY_FILE="/home/${SERVICE_USER}/.config/sops/age/keys.txt"
+if [ ! -f "${AGE_KEY_FILE}" ]; then
+    echo "ERROR: Age keypair not found at ${AGE_KEY_FILE}"
+    echo "Run infra/setup_server.sh first, then add the printed keys, then re-run."
+    exit 1
+fi
+
+# ── Clone or update repository ────────────────────────────────────────────────
+
+if [ -d "${REPO_DIR}/.git" ]; then
+    sudo -u "${SERVICE_USER}" git -C "${REPO_DIR}" fetch --tags --prune-tags origin
+else
+    sudo -u "${SERVICE_USER}" git clone \
+        "git@gitlab.com:${GITLAB_PROJECT}.git" "${REPO_DIR}"
+fi
+
+LATEST_TAG=$(sudo -u "${SERVICE_USER}" \
+    git -C "${REPO_DIR}" tag --list --sort=-version:refname "v*" | head -1)
+if [ -n "${LATEST_TAG}" ]; then
+    sudo -u "${SERVICE_USER}" git -C "${REPO_DIR}" checkout --detach "${LATEST_TAG}"
+fi
+
+# ── Decrypt secrets ───────────────────────────────────────────────────────────
+
+sudo -u "${SERVICE_USER}" bash -c \
+    "sops --input-type dotenv --output-type dotenv -d ${REPO_DIR}/.env.prod.sops > ${REPO_DIR}/.env"
+chmod 600 "${REPO_DIR}/.env"
+
+# ── Python dependencies ───────────────────────────────────────────────────────
+
+sudo -u "${SERVICE_USER}" bash -c "cd ${REPO_DIR} && ${UV} sync --all-packages"
+
+# ── Systemd services ──────────────────────────────────────────────────────────
+
+cp "${REPO_DIR}/infra/landing-backup/padelnomics-landing-backup.service" /etc/systemd/system/
+cp "${REPO_DIR}/infra/landing-backup/padelnomics-landing-backup.timer" /etc/systemd/system/
+cp "${REPO_DIR}/infra/supervisor/padelnomics-supervisor.service" /etc/systemd/system/
+systemctl daemon-reload
+systemctl enable --now padelnomics-landing-backup.timer
+systemctl enable --now padelnomics-supervisor
+
+echo ""
+echo "=== Bootstrap complete! ==="
+echo ""
+echo "Check status:   systemctl status padelnomics-supervisor"
+echo "View logs:      journalctl -u padelnomics-supervisor -f"
+echo "Backup timer:   systemctl list-timers padelnomics-landing-backup.timer"
+echo "Tag:            $(sudo -u "${SERVICE_USER}" git -C "${REPO_DIR}" describe --tags --always)"
--- a/infra/setup_server.sh
+++ b/infra/setup_server.sh
@@ -1,101 +1,176 @@
 #!/bin/bash
-# One-time server setup. Run as root on a fresh server.
-# Creates padelnomics_service user, installs system dependencies,
-# and registers systemd services that run as that user.
+# One-time server setup: create service user, install tools, SSH deploy key, age keypair.
+# Run as root on a fresh Debian/Ubuntu server before running bootstrap_supervisor.sh.
 #
 # Usage:
-#   sudo bash infra/setup_server.sh
+#   bash infra/setup_server.sh
+#
+# What it does:
+#   1. Creates padelnomics_service user (nologin) + adds to docker group
+#   2. Creates /opt/padelnomics + /data/padelnomics/landing with correct ownership
+#   3. Installs Docker, git, curl, age, sops, rclone, uv
+#   4. Generates ed25519 SSH deploy key for GitLab read access
+#   5. Generates age keypair at ~/.config/sops/age/keys.txt (as service user)
+#   6. Prints both public keys + numbered next-step instructions

 set -euo pipefail

-APP_DIR="/opt/padelnomics"
 SERVICE_USER="padelnomics_service"
-SERVICE_HOME="/home/${SERVICE_USER}"
-KEY_PATH="${SERVICE_HOME}/.ssh/padelnomics_deploy"
+APP_DIR="/opt/padelnomics"
+DATA_DIR="/data/padelnomics"
+SSH_DIR="/home/${SERVICE_USER}/.ssh"
+DEPLOY_KEY="${SSH_DIR}/padelnomics_deploy"
+SOPS_AGE_DIR="/home/${SERVICE_USER}/.config/sops/age"

-# Ensure running as root
-if [ "$(id -u)" -ne 0 ]; then
-    echo "Error: must run as root (use sudo)" >&2
-    exit 1
+[ "$(id -u)" = "0" ] || { echo "ERROR: Run as root: sudo bash infra/setup_server.sh"; exit 1; }
+
+log() { echo "$(date '+%H:%M:%S') ==> $*"; }
+
+# ── Service user ──────────────────────────────────────────────────────────────
+
+log "Creating service user ${SERVICE_USER}..."
+if ! id "${SERVICE_USER}" >/dev/null 2>&1; then
+    useradd --system --create-home --shell /usr/sbin/nologin "${SERVICE_USER}"
 fi

-# Create service user if not present
-if ! id "$SERVICE_USER" &>/dev/null; then
-    useradd --system --create-home --shell /usr/sbin/nologin "$SERVICE_USER"
-    echo "Created user $SERVICE_USER"
-else
-    echo "User $SERVICE_USER already exists, skipping"
+# ── Directories ───────────────────────────────────────────────────────────────
+
+log "Creating directories..."
+mkdir -p "${APP_DIR}" "${DATA_DIR}/landing"
+chown "${SERVICE_USER}:${SERVICE_USER}" "${APP_DIR}"
+chown -R "${SERVICE_USER}:${SERVICE_USER}" "${DATA_DIR}"
+
+# ── Docker ────────────────────────────────────────────────────────────────────
+
+if ! command -v docker >/dev/null 2>&1; then
+    log "Installing Docker..."
+    curl -fsSL https://get.docker.com | bash
+fi
+usermod -aG docker "${SERVICE_USER}"
+
+# ── System tools ──────────────────────────────────────────────────────────────
+
+log "Installing system tools..."
+apt-get update -q
+apt-get install -y -q git curl ca-certificates
+
+# ── SSH deploy key ────────────────────────────────────────────────────────────
+
+log "Setting up SSH deploy key..."
+sudo -u "${SERVICE_USER}" mkdir -p "${SSH_DIR}"
+chmod 700 "${SSH_DIR}"
+
+if [ ! -f "${DEPLOY_KEY}" ]; then
+    sudo -u "${SERVICE_USER}" ssh-keygen -t ed25519 \
+        -f "${DEPLOY_KEY}" -N "" -C "padelnomics-deploy"
 fi

-# Add service user to docker group (needed for deploy.sh)
-usermod -aG docker "$SERVICE_USER"
-echo "Added $SERVICE_USER to docker group"
-
-# Create app directory owned by service user
-mkdir -p "$APP_DIR"
-if [ "$(stat -c '%U' "$APP_DIR")" != "$SERVICE_USER" ]; then
-    chown "$SERVICE_USER:$SERVICE_USER" "$APP_DIR"
-fi
-echo "Created $APP_DIR"
-
-# Generate deploy key as service user if not present
-if [ ! -f "$KEY_PATH" ]; then
-    sudo -u "$SERVICE_USER" mkdir -p "${SERVICE_HOME}/.ssh"
-    sudo -u "$SERVICE_USER" ssh-keygen -t ed25519 -f "$KEY_PATH" -N "" -C "padelnomics-server"
-    chmod 700 "${SERVICE_HOME}/.ssh"
-    chmod 600 "$KEY_PATH"
-    chmod 644 "${KEY_PATH}.pub"
-
-    # Configure SSH to use this key for gitlab.com
-    if ! grep -q "# padelnomics" "${SERVICE_HOME}/.ssh/config" 2>/dev/null; then
-        sudo -u "$SERVICE_USER" tee -a "${SERVICE_HOME}/.ssh/config" > /dev/null <<EOF
-
-# padelnomics
+cat > "${SSH_DIR}/config" <<EOF
 Host gitlab.com
-    IdentityFile $KEY_PATH
+    IdentityFile ${DEPLOY_KEY}
+    IdentitiesOnly yes
 EOF
-        chmod 600 "${SERVICE_HOME}/.ssh/config"
-    fi
+chown "${SERVICE_USER}:${SERVICE_USER}" "${SSH_DIR}/config"
+chmod 600 "${SSH_DIR}/config"

-    echo "Generated deploy key: $KEY_PATH"
-else
-    echo "Deploy key already exists, skipping"
+ssh-keyscan -H gitlab.com >> "${SSH_DIR}/known_hosts" 2>/dev/null
+sort -u "${SSH_DIR}/known_hosts" -o "${SSH_DIR}/known_hosts"
+chown "${SERVICE_USER}:${SERVICE_USER}" "${SSH_DIR}/known_hosts"
+chmod 644 "${SSH_DIR}/known_hosts"
+
+# ── age ───────────────────────────────────────────────────────────────────────
+
+ARCH=$(uname -m)
+case "${ARCH}" in
+    x86_64)  ARCH_TAG="amd64" ;;
+    aarch64) ARCH_TAG="arm64" ;;
+    *) echo "Unsupported architecture: ${ARCH}"; exit 1 ;;
+esac
+
+if ! command -v age-keygen >/dev/null 2>&1; then
+    log "Installing age..."
+    AGE_VERSION="v1.3.1"
+    curl -fsSL "https://dl.filippo.io/age/${AGE_VERSION}?for=linux/${ARCH_TAG}" -o /tmp/age.tar.gz
+    tar -xzf /tmp/age.tar.gz -C /usr/local/bin --strip-components=1 age/age age/age-keygen
+    chmod +x /usr/local/bin/age /usr/local/bin/age-keygen
+    rm /tmp/age.tar.gz
 fi

-# Install rclone system-wide (we are root, no sudo needed)
-if ! command -v rclone &>/dev/null; then
-    echo "Installing rclone..."
+# ── sops ──────────────────────────────────────────────────────────────────────
+
+if ! command -v sops >/dev/null 2>&1; then
+    log "Installing sops..."
+    SOPS_VERSION="v3.12.1"
+    curl -fsSL \
+        "https://github.com/getsops/sops/releases/download/${SOPS_VERSION}/sops-${SOPS_VERSION}.linux.${ARCH_TAG}" \
+        -o /usr/local/bin/sops
+    chmod +x /usr/local/bin/sops
+fi
+
+# ── rclone ────────────────────────────────────────────────────────────────────
+
+if ! command -v rclone >/dev/null 2>&1; then
+    log "Installing rclone..."
    curl -fsSL https://rclone.org/install.sh | bash
-    echo "Installed rclone $(rclone --version | head -1)"
+fi
+
+# ── age keypair (as service user, XDG path auto-discovered by SOPS) ───────────
+
+log "Generating age keypair for ${SERVICE_USER}..."
+AGE_KEY_FILE="${SOPS_AGE_DIR}/keys.txt"
+if [ ! -f "${AGE_KEY_FILE}" ]; then
+    sudo -u "${SERVICE_USER}" mkdir -p "${SOPS_AGE_DIR}"
+    sudo -u "${SERVICE_USER}" age-keygen -o "${AGE_KEY_FILE}"
+    chmod 600 "${AGE_KEY_FILE}"
+    log "Age keypair generated at ${AGE_KEY_FILE}"
 else
-    echo "rclone already installed, skipping"
+    log "Age keypair already exists — skipping."
 fi

-# Create data directories owned by service user
-mkdir -p /data/padelnomics/landing
-if [ "$(stat -c '%U' /data/padelnomics)" != "$SERVICE_USER" ]; then
-    chown -R "$SERVICE_USER:$SERVICE_USER" /data/padelnomics
+# ── uv (installed as service user) ────────────────────────────────────────────
+
+if [ ! -f "/home/${SERVICE_USER}/.local/bin/uv" ]; then
+    log "Installing uv..."
+    sudo -u "${SERVICE_USER}" bash -c 'curl -LsSf https://astral.sh/uv/install.sh | sh'
 fi
-echo "Created /data/padelnomics/landing"

-# Install and enable systemd services
-cp "$APP_DIR/infra/landing-backup/padelnomics-landing-backup.service" /etc/systemd/system/
-cp "$APP_DIR/infra/landing-backup/padelnomics-landing-backup.timer" /etc/systemd/system/
-cp "$APP_DIR/infra/supervisor/padelnomics-supervisor.service" /etc/systemd/system/
-systemctl daemon-reload
-systemctl enable --now padelnomics-landing-backup.timer
-echo "Enabled landing backup timer (every 30 min)"
-systemctl enable --now padelnomics-supervisor.service
-echo "Enabled supervisor service"
+# ── Summary ───────────────────────────────────────────────────────────────────
+
+DEPLOY_PUB=$(cat "${DEPLOY_KEY}.pub")
+AGE_PUB=$(grep "public key:" "${AGE_KEY_FILE}" | awk '{print $NF}')

 echo ""
-echo "=== Next steps ==="
-echo "1. Add this deploy key to GitLab (Settings → Repository → Deploy Keys, read-only):"
+echo "=================================================================="
 echo ""
-cat "${KEY_PATH}.pub"
+echo "  SSH deploy key (add to GitLab → Settings → Deploy Keys):"
 echo ""
-echo "2. Clone the repo as $SERVICE_USER:"
-echo "   sudo -u $SERVICE_USER git clone git@gitlab.com:deemanone/padelnomics.git $APP_DIR"
+echo "    ${DEPLOY_PUB}"
 echo ""
-echo "3. Deploy (first run generates server age keypair — follow the printed instructions):"
-echo "   sudo -u $SERVICE_USER bash $APP_DIR/deploy.sh"
+echo "  Server age public key (add to .sops.yaml):"
+echo ""
+echo "    ${AGE_PUB}"
+echo ""
+echo "=================================================================="
+echo ""
+echo "  Next steps:"
+echo ""
+echo "  1. Add the SSH deploy key to GitLab:"
+echo "       → Repository Settings → Deploy Keys → Add key (read-only)"
+echo ""
+echo "  2. Add the age public key to .sops.yaml on your workstation:"
+echo "       creation_rules:"
+echo "         - path_regex: \\.env\\.(dev|prod)\\.sops\$"
+echo "           age: >-"
+echo "             <dev-key>"
+echo "             + ${AGE_PUB}"
+echo ""
+echo "  3. Re-encrypt prod secrets to include the server key:"
+echo "       sops updatekeys .env.prod.sops"
+echo "       git add .sops.yaml .env.prod.sops"
+echo "       git commit -m 'chore: add server age key'"
+echo "       git push"
+echo ""
+echo "  4. Run bootstrap:"
+echo "       bash infra/bootstrap_supervisor.sh"
+echo ""
+echo "=================================================================="