changemaker.lite/scripts/image-upgrade.sh
bunker-admin 4a3d9d7c41 feat(upgrade): Approach B - image-only upgrade mode
Add a "Quick Upgrade" path that pulls latest container images and recreates
only the core app services (api, admin, media-api, nginx) without touching
any tracked files. Tenant content (mkdocs/, configs/, scripts/) is implicitly
preserved because the script never writes outside docker.

Faster (~2 min vs ~4-5 min for full upgrade) and structurally safer for
releases that don't change orchestration/templates.

Pieces:
- scripts/image-upgrade.sh: new ~350-line script. Phases: pre-flight +
  mkdocs snapshot, image pull, targeted recreate (broad up -d would cascade
  on misconfigured infra containers — proven on marcelle), light health
  checks, deferred ccp-agent restart. Writes the same progress.json +
  result.json schema as upgrade.sh so the CCP poll loop is unchanged.
- agent/src/routes/upgrade.routes.ts: POST /instance/:slug/upgrade/start-image-only.
  Same lock + staleness guards as the existing /upgrade/start endpoint.
- api/src/services/remote-driver.ts: RemoteDriver.startImageUpgrade().
- api/src/services/upgrade.service.ts: startImageUpgrade() entry point;
  reuses runRemoteUpgrade with mode='image-only' (only the initial agent
  call differs — result schema and polling are identical).
- api/src/modules/instances/instances.routes.ts: POST /:id/upgrade-images
  + startImageUpgradeSchema.
- admin/src/pages/InstanceDetailPage.tsx: secondary "Quick Upgrade" button
  next to "Upgrade Now" on the Updates tab. Tooltip explains when to use it.

Tested locally on marcelle (v2.10.2 idempotent run): 1m 49s, mkdocs.yml md5
unchanged, file count unchanged, only api/admin/media-api/nginx touched.
Subtle bug found and fixed: `set -o pipefail` + `grep -q` shorts pipe and
SIGPIPEs the writer — captured services list once instead.

Bunker Admin
2026-05-21 15:20:35 -06:00

384 lines
14 KiB
Bash
Executable File

#!/usr/bin/env bash
# image-upgrade.sh — Approach B: image-only upgrade
#
# Pulls latest images from the registry and recreates services WITHOUT touching
# tracked files in the install tree (no git pull, no tarball extract, no VERSION
# mutation). Tenant content (mkdocs/, configs/) is implicitly safe because this
# script never writes outside data/upgrade/ and the docker daemon.
#
# Used by CCP "Quick Upgrade" button. Pairs with scripts/upgrade.sh which
# remains the full upgrade path for orchestration-changing releases.
#
# Schema parity: writes data/upgrade/progress.json + result.json with the same
# fields upgrade.sh writes, so the CCP poll loop is unchanged.
set -euo pipefail
PROJECT_DIR="$(cd "$(dirname "$(readlink -f "${BASH_SOURCE[0]}")")/.." && pwd)"
SCRIPT_DIR="$PROJECT_DIR/scripts"
UPGRADE_DIR="$PROJECT_DIR/data/upgrade"
LOG_DIR="$PROJECT_DIR/logs"
LOG_FILE="$LOG_DIR/image-upgrade-$(date +%Y%m%d_%H%M%S).log"
LOCK_FILE="$PROJECT_DIR/.upgrade.lock"
PROGRESS_FILE="$UPGRADE_DIR/progress.json"
RESULT_FILE="$UPGRADE_DIR/result.json"
START_TIME=$SECONDS
# --- Detect install mode ---
if [[ -f "$PROJECT_DIR/VERSION" ]] && [[ ! -d "$PROJECT_DIR/.git" ]]; then
INSTALL_MODE="release"
else
INSTALL_MODE="source"
fi
# --- Defaults ---
API_MODE=false
DRY_RUN=false
IMAGE_TAG=""
usage() {
cat <<EOF
Usage: $(basename "$0") [options]
Image-only upgrade: pulls latest images from the configured registry and
recreates services without touching the install tree.
Options:
--api-mode Emit data/upgrade/{progress,result}.json (no TTY output)
--dry-run Print what would happen; do not pull or recreate
--image-tag TAG Override IMAGE_TAG (env var) for this run
-h, --help Show this help
This script never modifies mkdocs/, configs/, scripts/, docker-compose.yml,
or VERSION. It is the safest upgrade path for orchestration-stable releases.
EOF
}
while [[ $# -gt 0 ]]; do
case "$1" in
--api-mode) API_MODE=true; shift ;;
--dry-run) DRY_RUN=true; shift ;;
--image-tag) IMAGE_TAG="${2:?--image-tag requires a value}"; shift 2 ;;
-h|--help) usage; exit 0 ;;
*) echo "Unknown option: $1" >&2; usage >&2; exit 1 ;;
esac
done
# --- Colors ---
if [[ -t 1 ]] && [[ -z "${NO_COLOR:-}" ]]; then
RED='\033[0;31m' GREEN='\033[0;32m' YELLOW='\033[0;33m'
CYAN='\033[0;36m' BOLD='\033[1m' NC='\033[0m'
else
RED='' GREEN='' YELLOW='' CYAN='' BOLD='' NC=''
fi
info() { echo -e "${CYAN}[INFO]${NC} $*"; }
success() { echo -e "${GREEN}[ OK ]${NC} $*"; }
warn() { echo -e "${YELLOW}[WARN]${NC} $*"; }
error() { echo -e "${RED}[ERR ]${NC} $*" >&2; }
phase() { echo ""; echo -e "${BOLD}${CYAN}=== Phase $1: $2 ===${NC}"; }
# --- Logging: mirror stdout/stderr to LOG_FILE ---
# logs/ may be root-owned on installs where upgrade.sh has run via ccp-agent.
# Fall back to /tmp if we can't write, so bunker-admin manual invocations don't
# crash with "Permission denied" on tee.
mkdir -p "$UPGRADE_DIR"
if mkdir -p "$LOG_DIR" 2>/dev/null && touch "$LOG_FILE" 2>/dev/null; then
: # primary log location is writable
else
LOG_FILE="/tmp/image-upgrade-$(date +%Y%m%d_%H%M%S)-$$.log"
echo "[INFO] logs/ not writable; using $LOG_FILE" >&2
fi
exec > >(tee -a "$LOG_FILE") 2>&1
# --- Capture previous version for result.json ---
if [[ "$INSTALL_MODE" == "release" ]]; then
PRE_VERSION="$(head -1 "$PROJECT_DIR/VERSION" 2>/dev/null || echo "unknown")"
else
PRE_VERSION="$(cd "$PROJECT_DIR" && git rev-parse --short HEAD 2>/dev/null || echo "unknown")"
fi
write_progress() {
local phase_num="$1" phase_name="$2" pct="$3" msg="$4"
[[ "$API_MODE" != "true" ]] && return
mkdir -p "$UPGRADE_DIR"
cat > "$PROGRESS_FILE" <<PEOF
{
"phase": ${phase_num},
"phaseName": "${phase_name}",
"percentage": ${pct},
"message": "$(echo "$msg" | sed 's/"/\\"/g')",
"lastUpdate": "$(date -u +%Y-%m-%dT%H:%M:%SZ)"
}
PEOF
}
write_result() {
[[ "$API_MODE" != "true" ]] && return
local success_val="$1" msg="$2"
local warnings_json="${3:-[]}"
local duration_secs=$((SECONDS - START_TIME))
local new_version="$PRE_VERSION"
if [[ "$INSTALL_MODE" == "release" ]]; then
new_version="$(head -1 "$PROJECT_DIR/VERSION" 2>/dev/null || echo "$PRE_VERSION")"
else
new_version="$(cd "$PROJECT_DIR" && git rev-parse --short HEAD 2>/dev/null || echo "$PRE_VERSION")"
fi
mkdir -p "$UPGRADE_DIR"
cat > "$RESULT_FILE" <<REOF
{
"success": ${success_val},
"message": "$(echo "$msg" | sed 's/"/\\"/g')",
"previousCommit": "${PRE_VERSION}",
"newCommit": "${new_version}",
"commitCount": 0,
"durationSeconds": ${duration_secs},
"warnings": ${warnings_json},
"completedAt": "$(date -u +%Y-%m-%dT%H:%M:%SZ)",
"mode": "image-only"
}
REOF
rm -f "$PROGRESS_FILE"
}
# --- Lock + cleanup ---
acquire_lock() {
if [[ -f "$LOCK_FILE" ]]; then
local pid; pid="$(cat "$LOCK_FILE" 2>/dev/null || echo "")"
if [[ -n "$pid" ]] && kill -0 "$pid" 2>/dev/null; then
error "Upgrade already running (pid $pid). Refusing to start."
write_result "false" "Another upgrade is already running (pid $pid)"
exit 1
fi
warn "Stale lock file found; removing"
rm -f "$LOCK_FILE"
fi
echo $$ > "$LOCK_FILE"
}
release_lock() { rm -f "$LOCK_FILE" || true; }
on_failure() {
local exit_code=$?
local line_no=${1:-?}
error "image-upgrade.sh failed at line $line_no (exit $exit_code)"
write_result "false" "Image upgrade failed at line $line_no (exit $exit_code)"
release_lock
exit "$exit_code"
}
trap 'on_failure $LINENO' ERR
trap 'release_lock' EXIT
# --- Banner ---
echo ""
echo -e "${BOLD}${CYAN}================================================${NC}"
echo -e "${BOLD} Image-Only Upgrade${NC}"
echo -e "${BOLD}${CYAN}================================================${NC}"
echo "Install mode: $INSTALL_MODE"
echo "Project dir: $PROJECT_DIR"
echo "Pre-version: $PRE_VERSION"
[[ -n "$IMAGE_TAG" ]] && echo "Image tag: $IMAGE_TAG"
[[ "$DRY_RUN" == "true" ]] && echo "DRY RUN: no images will be pulled or services recreated"
echo ""
acquire_lock
# =============================================================================
# Phase 1: Pre-flight + mkdocs snapshot (defensive)
# =============================================================================
phase "1" "Pre-flight"
write_progress 1 "Pre-flight" 10 "Snapshotting mkdocs (defensive)..."
# Source mkdocs-snapshot.sh and run it. This is the same snapshot every
# upgrade path takes — leaves mkdocs-backup-<timestamp>.tar.gz in project root.
# Image-only upgrades shouldn't damage mkdocs (no filesystem mutation), but
# the snapshot is cheap insurance and keeps operator habits consistent.
if [[ -r "$SCRIPT_DIR/lib/mkdocs-snapshot.sh" ]]; then
if [[ "$DRY_RUN" == "true" ]]; then
info "[DRY RUN] Would snapshot mkdocs/"
else
# shellcheck disable=SC1091
PROJECT_DIR="$PROJECT_DIR" bash -c ". $SCRIPT_DIR/lib/mkdocs-snapshot.sh; snapshot_mkdocs" \
|| warn "mkdocs snapshot failed (non-fatal; continuing)"
fi
else
warn "scripts/lib/mkdocs-snapshot.sh not found; skipping snapshot"
fi
# Sanity-check docker
if ! docker compose version &>/dev/null; then
error "docker compose is not available"
write_result "false" "docker compose not available"
exit 1
fi
success "Pre-flight checks passed"
# =============================================================================
# Phase 2: Pull images
# =============================================================================
phase "2" "Pull Images"
write_progress 2 "Pull Images" 30 "Pulling images from registry..."
PULL_ENV=()
if [[ -n "$IMAGE_TAG" ]]; then
PULL_ENV+=("IMAGE_TAG=$IMAGE_TAG")
fi
if [[ "$DRY_RUN" == "true" ]]; then
info "[DRY RUN] Would run: ${PULL_ENV[*]:-} docker compose pull"
else
info "Pulling all images (this may take a few minutes)..."
if (( ${#PULL_ENV[@]} > 0 )); then
if ! env "${PULL_ENV[@]}" docker compose pull; then
warn "docker compose pull had errors (continuing — some images may be local)"
fi
else
if ! docker compose pull; then
warn "docker compose pull had errors (continuing — some images may be local)"
fi
fi
fi
success "Image pull complete"
# =============================================================================
# Phase 3: Recreate core app services (targeted, not broad)
# =============================================================================
phase "3" "Recreate Services"
write_progress 3 "Recreate Services" 60 "Recreating core app services with new images..."
# Targeted recreate: only the services whose IMAGES are released as part of
# changemaker.lite (api, admin, media-api, nginx). Broader `up -d` is risky
# because a single misconfigured mount in any service (e.g. mkdocs-site-server)
# can cascade and leave dependent containers in "Created" state. Image-only
# upgrade should only touch the actual code containers, not third-party
# infrastructure that happens to live in the same compose file.
#
# Same Phase 6 pattern as upgrade.sh: drop ccp-agent from COMPOSE_PROFILES
# during recreate so we don't suicide-restart the agent that spawned us.
# Restart ccp-agent at the end via detached subshell.
PROFILES_SAVED="${COMPOSE_PROFILES:-}"
COMPOSE_PROFILES_WITHOUT_AGENT="$(echo "${PROFILES_SAVED}" \
| tr ',' '\n' | grep -vx 'ccp-agent' | paste -sd, -)"
UP_ENV=("COMPOSE_PROFILES=${COMPOSE_PROFILES_WITHOUT_AGENT}")
if [[ -n "$IMAGE_TAG" ]]; then
UP_ENV+=("IMAGE_TAG=$IMAGE_TAG")
fi
# Core services that ship as v2 release images. nginx last so it doesn't
# briefly proxy to an old api. media-api may not be enabled on all installs;
# tolerate it being missing from compose.
CORE_SERVICES=(api admin media-api nginx)
EXISTING_SERVICES=()
# Capture the service list once. Don't pipe `docker compose config` into
# `grep -q` directly: with `set -o pipefail`, grep exits early on match and
# SIGPIPEs the docker writer, making the pipeline exit non-zero. The grep -q
# would then "match" all services as missing. Capture-then-check avoids it.
COMPOSE_SERVICES_LIST="$(docker compose config --services 2>/dev/null || true)"
for svc in "${CORE_SERVICES[@]}"; do
if grep -qx -- "$svc" <<<"$COMPOSE_SERVICES_LIST"; then
EXISTING_SERVICES+=("$svc")
else
info "Skipping service '$svc' (not in compose file)"
fi
done
if (( ${#EXISTING_SERVICES[@]} == 0 )); then
warn "No core app services found in compose; skipping recreate"
elif [[ "$DRY_RUN" == "true" ]]; then
info "[DRY RUN] Would run: ${UP_ENV[*]} docker compose up -d ${EXISTING_SERVICES[*]}"
else
info "Recreating core services: ${EXISTING_SERVICES[*]}"
env "${UP_ENV[@]}" docker compose up -d "${EXISTING_SERVICES[@]}"
fi
success "Services recreated"
# Restart Pangolin tunnel connector if running (image may have changed)
if docker ps --format '{{.Names}}' | grep -q 'newt'; then
if [[ "$DRY_RUN" == "true" ]]; then
info "[DRY RUN] Would restart newt"
else
info "Restarting Pangolin tunnel connector..."
docker compose restart newt 2>/dev/null || true
success "Newt tunnel restarted"
fi
fi
# =============================================================================
# Phase 4: Verify (light health checks)
# =============================================================================
phase "4" "Verification"
write_progress 4 "Verification" 85 "Running health checks..."
VERIFY_FAILED=false
UPGRADE_WARNINGS="[]"
verify_health() {
local name="$1" check_cmd="$2" max_wait="${3:-45}"
local waited=0
while [[ $waited -lt $max_wait ]]; do
if eval "$check_cmd" 2>/dev/null; then
success "$name: healthy (${waited}s)"
return 0
fi
sleep 3
waited=$((waited + 3))
done
warn "$name: not responding after ${max_wait}s"
VERIFY_FAILED=true
return 0
}
if [[ "$DRY_RUN" != "true" ]]; then
verify_health "API (port 4000)" \
"docker compose exec -T api wget -q --spider http://localhost:4000/api/health" 60
verify_health "Admin (port 3000)" \
"docker compose exec -T admin wget -q --spider http://localhost:3000/" 90
if docker ps --format '{{.Names}}' | grep -q 'changemaker-media-api'; then
verify_health "Media API (port 4100)" \
"docker compose exec -T media-api wget -q --spider http://127.0.0.1:4100/health" 30
fi
if "$VERIFY_FAILED"; then
UPGRADE_WARNINGS='["Some health checks failed after image-only upgrade — services may still be starting"]'
fi
fi
# =============================================================================
# Summary + deferred ccp-agent restart
# =============================================================================
ELAPSED_MIN=$(( (SECONDS - START_TIME) / 60 ))
ELAPSED_SEC=$(( (SECONDS - START_TIME) % 60 ))
echo ""
echo -e "${BOLD}${GREEN}================================================${NC}"
echo -e "${BOLD} Image-Only Upgrade Complete${NC}"
echo -e "${BOLD}${GREEN}================================================${NC}"
printf " Previous: %s\n" "$PRE_VERSION"
printf " Duration: %dm %ds\n" "$ELAPSED_MIN" "$ELAPSED_SEC"
printf " Log: %s\n" "$LOG_FILE"
write_progress 4 "Complete" 100 "Image-only upgrade complete"
write_result "true" "Image-only upgrade complete (previous: ${PRE_VERSION})" "$UPGRADE_WARNINGS"
# Deferred ccp-agent restart — see upgrade.sh for full rationale. Same
# mechanism: nohup'd, disowned subshell that picks up the new image after
# this script has cleanly exited.
if echo "${PROFILES_SAVED:-}" | tr ',' '\n' | grep -qx 'ccp-agent'; then
if [[ "$DRY_RUN" == "true" ]]; then
info "[DRY RUN] Would schedule deferred ccp-agent restart"
else
info "Scheduling deferred ccp-agent restart..."
nohup bash -c "
sleep 3
cd '$PROJECT_DIR'
COMPOSE_PROFILES='ccp-agent' docker compose --profile ccp-agent up -d ccp-agent
" >/dev/null 2>&1 < /dev/null &
disown
success "ccp-agent restart scheduled (will pick up new image)"
fi
fi
release_lock
trap - EXIT
exit 0