feat: harden multi-host runtime coordination

src/govoplan_core/celery_app.py

@@ -3,9 +3,15 @@ from __future__ import annotations
 from datetime import datetime, timedelta, timezone
 from importlib.metadata import PackageNotFoundError, version
 import logging
+import os
 
 from celery import Celery
-from celery.signals import heartbeat_sent, worker_ready, worker_shutdown
+from celery.signals import (
+    heartbeat_sent,
+    worker_process_init,
+    worker_ready,
+    worker_shutdown,
+)
 
 from govoplan_core.core.campaigns import (
     CAPABILITY_CAMPAIGNS_DELIVERY_TASKS,
@@ -196,6 +202,21 @@ def _worker_runtime_identity(sender: object | None = None) -> RuntimeIdentity:
     return _worker_identity
 
 
+def _worker_metadata() -> dict[str, object]:
+    raw_concurrency = str(os.getenv("CELERY_WORKER_CONCURRENCY") or "").strip()
+    return {
+        "process": "celery-worker",
+        "worker_pool": str(os.getenv("GOVOPLAN_WORKER_POOL") or "default"),
+        "concurrency": int(raw_concurrency) if raw_concurrency.isdigit() else None,
+    }
+
+
+@worker_process_init.connect
+def _reset_worker_process_database(**_kwargs) -> None:
+    # SQLAlchemy pools must not be shared across prefork child processes.
+    configure_database(settings.database_url, dispose_previous=True)
+
+
 @worker_ready.connect
 def _register_worker_runtime(sender=None, **_kwargs) -> None:
     global _worker_consumer
@@ -206,7 +227,7 @@ def _register_worker_runtime(sender=None, **_kwargs) -> None:
             node = register_runtime_node(
                 session,
                 identity,
-                metadata={"process": "celery-worker"},
+                metadata=_worker_metadata(),
             )
             session.commit()
     except Exception:
@@ -227,7 +248,7 @@ def _heartbeat_worker_runtime(sender=None, **_kwargs) -> None:
             node = heartbeat_runtime_node(
                 session,
                 identity,
-                metadata={"process": "celery-worker"},
+                metadata=_worker_metadata(),
             )
             session.commit()
     except Exception:  # noqa: BLE001 - uncertain authority must fail closed