Improve Local Agent update reliability v1.26.24

app/modules/documents/local_agent_runtime/ERPAgentSupervisor.pyw

@@ -23,6 +23,7 @@ STATE_FILE = DATA_DIR / "supervisor_state.json"
 OWNER_FILE = DATA_DIR / "supervisor_owner.json"
 REQUEST_FILE = UPDATES_DIR / "supervisor_request.json"
 UPDATE_PROGRESS_FILE = DATA_DIR / "update_progress.json"
+FAILED_UPDATES_FILE = DATA_DIR / "failed_updates.json"
 WORKER_LOG = LOG_DIR / "worker-supervisor.log"
 SUPERVISOR_LOG = LOG_DIR / "supervisor.log"
 DASHBOARD_URL = "http://127.0.0.1:8788"
@@ -110,6 +111,36 @@ def _update_progress(phase: str, percent: int, message: str, *, version: str = "
     _write_json_atomic(UPDATE_PROGRESS_FILE, payload)
 
 
+def _read_failed_updates() -> dict:
+    try:
+        if FAILED_UPDATES_FILE.exists():
+            value = json.loads(FAILED_UPDATES_FILE.read_text(encoding="utf-8"))
+            return value if isinstance(value, dict) else {}
+    except Exception:
+        pass
+    return {}
+
+
+def _mark_failed_update(version: str, error: str) -> None:
+    if not version:
+        return
+    payload = _read_failed_updates()
+    previous = payload.get(version) if isinstance(payload.get(version), dict) else {}
+    payload[version] = {
+        "failed_at": time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime()),
+        "error": str(error)[:2000],
+        "attempts": int(previous.get("attempts") or 0) + 1,
+    }
+    _write_json_atomic(FAILED_UPDATES_FILE, payload)
+
+
+def _clear_failed_update(version: str) -> None:
+    payload = _read_failed_updates()
+    if version in payload:
+        payload.pop(version, None)
+        _write_json_atomic(FAILED_UPDATES_FILE, payload)
+
+
 def _dashboard_ready(timeout: float = 1.5) -> bool:
     try:
         with urllib.request.urlopen(DASHBOARD_URL + "/api/status", timeout=timeout) as response:
@@ -443,6 +474,7 @@ class Supervisor:
                     raise RuntimeError(f"Updated worker exited with code {self.worker.returncode}.")
                 if _dashboard_ready():
                     _update_progress("complete", 100, f"ERP Local Agent {version} updated successfully.", version=version, status="updated")
+                    _clear_failed_update(version)
                     _state(
                         "updated",
                         f"ERP Local Agent {version} installed and restarted successfully.",
@@ -456,6 +488,7 @@ class Supervisor:
                 time.sleep(1)
             raise RuntimeError("Updated worker did not become healthy within the restart timeout.")
         except Exception as exc:
+            _mark_failed_update(version, str(exc))
             _update_progress("rollback", 92, f"Update failed; restoring the previous runtime: {exc}", version=version, status="rollback")
             _state("rollback", f"Update {version} failed. Restoring previous worker.", target_version=version)
             self.stop_worker()