fix: revert to robust manual leadership detection to prevent multiple Navidrome instances

entrypoint.sh

@@ -7,12 +7,49 @@ SERVICE_ID="${SERVICE_NAME}-${NOMAD_ALLOC_ID:-$(hostname)}"
 PORT=4533
 CONSUL_HTTP_ADDR="${CONSUL_URL:-http://localhost:8500}"
 NODE_IP="${ADVERTISE_IP}"
+DB_LOCK_FILE="/litefs/.primary"
+NAVIDROME_PID=0
 
 # Tags for the Primary service (Traefik enabled)
 PRIMARY_TAGS='["navidrome","web","traefik.enable=true","urlprefix-/navidrome","tools","traefik.http.routers.navidromelan.rule=Host(`navidrome.service.dc1.consul`)","traefik.http.routers.navidromewan.rule=Host(`m.fbleagh.duckdns.org`)","traefik.http.routers.navidromewan.middlewares=dex@consulcatalog","traefik.http.routers.navidromewan.tls=true"]'
 
 # --- Helper Functions ---
 
+# Check if this node is the LiteFS Primary
+# LiteFS 0.5 status API returns a flat object: {"isPrimary": true, ...}
+check_primary() {
+  local status=$(curl -s http://localhost:20202/status || echo "{}")
+  local is_primary=$(echo "$status" | jq -r 'if type == "object" then (.isPrimary // false) else false end' 2>/dev/null || echo "false")
+  
+  if [ "$is_primary" = "true" ]; then
+    return 0 # We are the primary
+  fi
+  return 1 # We are a replica
+}
+
+# Wait for LiteFS to settle and determine its role
+wait_for_litefs() {
+  echo "Waiting for LiteFS to settle..."
+  local timeout=60
+  local count=0
+  while [ $count -lt $timeout ]; do
+    local status=$(curl -s http://localhost:20202/status || echo "null")
+    local is_primary_val=$(echo "$status" | jq -r 'if type == "object" then (.isPrimary // "null") else "null" end' 2>/dev/null || echo "null")
+    
+    if [ "$is_primary_val" != "null" ]; then
+      local role="replica"
+      if [ "$is_primary_val" = "true" ]; then role="primary"; fi
+      echo "LiteFS initialized. Role: $role"
+      return 0
+    fi
+    sleep 2
+    count=$((count + 2))
+    echo -n "."
+  done
+  echo "ERROR: LiteFS failed to settle after ${timeout}s"
+  return 1
+}
+
 # Register Service with TTL Check
 register_service() {
   echo "Registering service ${SERVICE_ID} with Consul..."
@@ -40,37 +77,61 @@ deregister_service() {
   curl -s -X PUT "${CONSUL_HTTP_ADDR}/v1/agent/service/deregister/${SERVICE_ID}"
 }
 
+# Start Navidrome in Background
+start_app() {
+  echo "Node is Primary. Starting Navidrome..."
+  
+  # Ensure shared directories exist on persistent host volume
+  mkdir -p /data/plugins /data/cache /data/backup
+
+  # Tell Navidrome to use the database directly from the LiteFS mount.
+  export ND_DBPATH="/litefs/navidrome.db?cache=shared&_busy_timeout=15000&_journal_mode=WAL&_foreign_keys=on"
+
+  /app/navidrome &
+  NAVIDROME_PID=$!
+  echo "Navidrome running (PID: $NAVIDROME_PID)"
+}
+
+# Stop Navidrome
+stop_app() {
+  if [ "${NAVIDROME_PID}" -gt 0 ]; then
+    echo "Stopping Navidrome (PID ${NAVIDROME_PID})..."
+    kill -SIGTERM "${NAVIDROME_PID}"
+    wait "${NAVIDROME_PID}" 2>/dev/null || true
+    NAVIDROME_PID=0
+  fi
+}
+
 # --- Cleanup ---
 cleanup() {
   echo "Shutting down..."
+  stop_app
   deregister_service
   exit 0
 }
 
 trap cleanup TERM INT
 
-# --- Main Logic ---
+# --- Main Loop ---
 
-echo "Starting Navidrome Primary Node..."
+echo "Starting Supervisor. Waiting for leadership settle..."
+wait_for_litefs || exit 1
 
-# 1. Ensure shared directories exist on persistent host volume
-mkdir -p /data/plugins /data/cache /data/backup
-
-# 2. Tell Navidrome to use the database directly from the LiteFS mount.
-export ND_DBPATH="/litefs/navidrome.db?cache=shared&_busy_timeout=15000&_journal_mode=WAL&_foreign_keys=on"
-
-# 3. Register with Consul
-register_service
-
-# 4. Start Navidrome and Maintain TTL
-/app/navidrome &
-NAVIDROME_PID=$!
-
-echo "Navidrome running (PID: $NAVIDROME_PID)"
-
-while kill -0 $NAVIDROME_PID 2>/dev/null; do
-  pass_ttl
+while true; do
+  if check_primary; then
+    # === WE ARE PRIMARY ===
+    if [ "${NAVIDROME_PID}" -eq 0 ] || ! kill -0 "${NAVIDROME_PID}" 2>/dev/null; then
+      start_app
+      register_service
+    fi
+    pass_ttl
+  else
+    # === WE ARE REPLICA ===
+    if [ "${NAVIDROME_PID}" -gt 0 ]; then
+      echo "Lost leadership. Demoting..."
+      stop_app
+      deregister_service
+    fi
+  fi
   sleep 10
 done
-
-cleanup

litefs.yml

@@ -36,4 +36,3 @@ proxy:
 # Commands to run only on the primary node. 
 exec:
   - cmd: "/usr/local/bin/entrypoint.sh"
-    if-candidate: true