feat: add deployment configuration and scripts for managed-portal, including Dockerfiles and environment settings

managed/people_flow_project/scripts/docker-entrypoint.sh

@@ -8,8 +8,15 @@ OUTPUT_DIR="${OUTPUT_DIR:-${PROJECT_DIR}/outputs}"
 RTSP_URL="${RTSP_URL:-}"
 API_HOST="${API_HOST:-0.0.0.0}"
 API_PORT="${API_PORT:-18082}"
+RTSP_STALL_TIMEOUT_SECONDS="${RTSP_STALL_TIMEOUT_SECONDS:-180}"
+DEEPFACE_CACHE_DIR="/root/.deepface/weights"
+DEEPFACE_SOURCE_DIR="${PROJECT_DIR}/weights/deepface"
 
-mkdir -p "${OUTPUT_DIR}" "$(dirname "${CONFIG_PATH}")"
+mkdir -p "${OUTPUT_DIR}" "$(dirname "${CONFIG_PATH}")" "${DEEPFACE_CACHE_DIR}"
+
+if [ -d "${DEEPFACE_SOURCE_DIR}" ]; then
+  find "${DEEPFACE_SOURCE_DIR}" -maxdepth 1 -name '*.h5' -exec cp {} "${DEEPFACE_CACHE_DIR}/" \;
+fi
 
 if [ ! -f "${CONFIG_PATH}" ]; then
   cp "${CONFIG_TEMPLATE}" "${CONFIG_PATH}"
@@ -37,13 +44,31 @@ config_path.write_text(
 )
 PY
 
-exec python - "$CONFIG_PATH" "$API_HOST" "$API_PORT" <<'PY'
+RTSP_OUTPUT_SUBDIR="$(python - "$CONFIG_PATH" <<'PY'
+from pathlib import Path
+import sys
+import yaml
+
+config_path = Path(sys.argv[1])
+raw = yaml.safe_load(config_path.read_text(encoding="utf-8")) or {}
+rtsp = raw.get("rtsp") or {}
+print(rtsp.get("output_subdir", "rtsp_stream"))
+PY
+)"
+RTSP_STATUS_PATH="${OUTPUT_DIR}/${RTSP_OUTPUT_SUBDIR}/worker_status.json"
+
+exec python - "$CONFIG_PATH" "$API_HOST" "$API_PORT" "$RTSP_STATUS_PATH" "$RTSP_STALL_TIMEOUT_SECONDS" <<'PY'
+from pathlib import Path
 import signal
 import subprocess
 import sys
 import time
 
-config_path, api_host, api_port = sys.argv[1:4]
+from src.people_flow.worker_status import worker_status_stall_reason
+
+config_path, api_host, api_port, status_path_raw, stall_timeout_raw = sys.argv[1:6]
+status_path = Path(status_path_raw)
+stall_timeout_seconds = max(float(stall_timeout_raw), 30.0)
 commands = [
   [sys.executable, "main.py", "--config", config_path, "rtsp"],
   [
@@ -59,21 +84,38 @@ commands = [
   ],
 ]
 processes = [subprocess.Popen(command) for command in commands]
+supervisor_started_at = time.time()
 
 
-def terminate_all(signum, _frame):
-  for process in processes:
-    if process.poll() is None:
-      process.terminate()
+def stop_all(excluded_index=None):
+  for index, process in enumerate(processes):
+    if index == excluded_index or process.poll() is not None:
+      continue
+    process.terminate()
   deadline = time.time() + 10
-  for process in processes:
-    if process.poll() is not None:
+  for index, process in enumerate(processes):
+    if index == excluded_index or process.poll() is not None:
       continue
     timeout = max(0, deadline - time.time())
     try:
       process.wait(timeout=timeout)
     except subprocess.TimeoutExpired:
       process.kill()
+
+
+def stale_reason():
+  if processes[0].poll() is not None:
+    return None
+
+  return worker_status_stall_reason(
+    status_path,
+    started_at=supervisor_started_at,
+    max_age_seconds=stall_timeout_seconds,
+  )
+
+
+def terminate_all(signum, _frame):
+  stop_all()
   raise SystemExit(128 + signum)
 
 
@@ -85,19 +127,12 @@ while True:
     return_code = process.poll()
     if return_code is None:
       continue
-    for other_index, other_process in enumerate(processes):
-      if other_index == index or other_process.poll() is not None:
-        continue
-      other_process.terminate()
-    deadline = time.time() + 10
-    for other_index, other_process in enumerate(processes):
-      if other_index == index or other_process.poll() is not None:
-        continue
-      timeout = max(0, deadline - time.time())
-      try:
-        other_process.wait(timeout=timeout)
-      except subprocess.TimeoutExpired:
-        other_process.kill()
+    stop_all(excluded_index=index)
     raise SystemExit(return_code)
+  reason = stale_reason()
+  if reason is not None:
+    print(reason, flush=True)
+    stop_all()
+    raise SystemExit(1)
   time.sleep(0.5)
 PY