fix(gateway): flush pending messages to disk before shutdown clear (#72680)

When FTS5 index corruption prevents INSERT INTO messages, the gateway
accumulates messages in _pending_messages (memory-only). On shutdown,
.clear() discards the only surviving copy — permanent user data loss.

Changes:
- Add gateway/shutdown_flush.py with flush_pending_to_file() and
  recover_pending_to_db() for two-phase data preservation.
- Patch gateway/run.py: flush runner._pending_messages before clear()
  in _stop_impl_body, and recover on startup after runner.start().
- Patch gateway/platforms/base.py: flush adapter._pending_messages
  before clear() in the adapter shutdown path.

Recovery behavior:
- Reads pending JSON files from ~/.hermes/pending_messages/
- Inserts messages into state.db directly
- Per-session isolation: one corrupt session doesn't block others
- Successful recovery deletes the flush file
- Failed recovery re-saves for next startup retry
This commit is contained in:
JonthanaHanh 2026-07-28 08:09:17 +07:00 committed by kshitij
parent f134f8cac6
commit 58f6678e6d
3 changed files with 220 additions and 0 deletions

View file

@ -6073,6 +6073,12 @@ class BasePlatformAdapter(ABC):
self._background_tasks.clear()
self._expected_cancelled_tasks.clear()
self._session_tasks.clear()
# Flush pending messages to disk before clearing (#72680).
try:
from gateway.shutdown_flush import flush_pending_to_file
flush_pending_to_file(self._pending_messages, reason="adapter_shutdown")
except Exception:
pass
self._pending_messages.clear()
self._active_sessions.clear()
for state in list(self._text_debounce_store().values()):

View file

@ -9968,6 +9968,15 @@ class GatewayRunner(GatewayAuthorizationMixin, GatewayKanbanWatchersMixin, Gatew
self._running_agents_ts.clear()
if hasattr(self, "_active_session_leases"):
self._active_session_leases.clear()
# Flush pending messages to disk before clearing (#72680).
# When FTS5 corruption prevents message persistence, the
# in-memory _pending_messages dict holds the only surviving
# copy. Clearing without flushing causes permanent data loss.
try:
from gateway.shutdown_flush import flush_pending_to_file
flush_pending_to_file(self._pending_messages, reason="shutdown")
except Exception:
pass
self._pending_messages.clear()
self._pending_approvals.clear()
if hasattr(self, '_busy_ack_ts'):
@ -24490,6 +24499,16 @@ async def start_gateway(config: Optional[GatewayConfig] = None, replace: bool =
success = await runner.start()
if not success:
return False
# Recover any pending messages flushed during a previous shutdown (#72680).
try:
from gateway.shutdown_flush import recover_pending_to_db
recovered = recover_pending_to_db()
if recovered:
logger.info(
"Recovered %d pending message(s) from shutdown flush", recovered,
)
except Exception:
pass
if runner.should_exit_cleanly:
if runner.exit_reason:
logger.error("Gateway exiting cleanly: %s", runner.exit_reason)

195
gateway/shutdown_flush.py Normal file
View file

@ -0,0 +1,195 @@
"""Flush pending messages to disk before shutdown to prevent data loss.
When FTS5 index corruption prevents ``INSERT INTO messages``, the gateway
accumulates messages in ``_pending_messages`` (memory-only). On shutdown,
``.clear()`` discards the only surviving copy permanent user data loss.
This module provides two hooks:
1. ``flush_pending_to_file()`` called BEFORE ``_pending_messages.clear()``
during shutdown. Serialises any non-empty pending slots to a JSON file
under ``~/.hermes/pending_messages/``.
2. ``recover_pending_to_db()`` called AFTER ``runner.start()`` on startup.
Reads flush files, inserts messages directly into state.db, then deletes
the flush file on success.
See issue #72680 for the full incident report.
"""
from __future__ import annotations
import json
import logging
import time
from pathlib import Path
from typing import Any, Dict, Optional
logger = logging.getLogger(__name__)
_FLUSH_DIR = Path.home() / ".hermes" / "pending_messages"
def _get_flush_dir() -> Path:
_FLUSH_DIR.mkdir(parents=True, exist_ok=True)
return _FLUSH_DIR
def flush_pending_to_file(
pending: Dict[str, Any],
*,
reason: str = "shutdown",
) -> int:
"""Serialise non-empty ``_pending_messages`` slots to disk.
Parameters
----------
pending:
The adapter or runner ``_pending_messages`` dict. Values may be
``MessageEvent`` objects (adapter) or plain strings (runner).
reason:
Logged context (``shutdown``, ``restart``, etc.).
Returns
-------
int
Number of sessions flushed.
"""
if not pending:
return 0
flush_dir = _get_flush_dir()
ts = int(time.time())
flushed = 0
for session_key, value in list(pending.items()):
if value is None:
continue
try:
serialised = _serialise_value(value)
if serialised is None:
continue
safe_key = session_key.replace("/", "_").replace("\\", "_")
path = flush_dir / f"{safe_key}_{ts}.json"
path.write_text(
json.dumps({
"session_key": session_key,
"reason": reason,
"ts": ts,
"data": serialised,
}, ensure_ascii=False, default=str),
encoding="utf-8",
)
flushed += 1
except Exception as exc:
logger.debug(
"Failed to flush pending message for %s: %s",
session_key, exc,
)
if flushed:
logger.info(
"Flushed %d pending message(s) to %s (reason=%s)",
flushed, flush_dir, reason,
)
return flushed
def _serialise_value(value: Any) -> Optional[dict]:
"""Convert a pending message value to a JSON-serialisable dict."""
# MessageEvent objects have a .text attribute and other fields
if hasattr(value, "text"):
result: Dict[str, Any] = {"text": getattr(value, "text", "")}
# Preserve additional fields if present
for attr in ("session_id", "platform", "sender_id", "sender_name",
"reply_to", "media", "raw_event"):
val = getattr(value, attr, None)
if val is not None:
try:
json.dumps(val)
result[attr] = val
except (TypeError, ValueError):
result[attr] = str(val)
return result
# Plain string (runner-level _pending_messages)
if isinstance(value, str):
return {"text": value}
# Dict — try direct serialisation
if isinstance(value, dict):
try:
json.dumps(value)
return value
except (TypeError, ValueError):
return {"text": str(value)}
return {"text": str(value)}
def recover_pending_to_db(
db_path: Optional[Path] = None,
) -> int:
"""Recover flushed pending messages into state.db.
Reads all ``*.json`` files from the flush directory, inserts messages
into the ``messages`` table, and deletes the file on success.
Parameters
----------
db_path:
Path to state.db. Defaults to ``~/.hermes/state.db``.
Returns
-------
int
Number of messages recovered.
"""
import sqlite3
flush_dir = _get_flush_dir()
flush_files = sorted(flush_dir.glob("*.json"))
if not flush_files:
return 0
if db_path is None:
db_path = Path.home() / ".hermes" / "state.db"
if not db_path.exists():
logger.warning("state.db not found at %s — skipping recovery", db_path)
return 0
recovered = 0
for path in flush_files:
try:
payload = json.loads(path.read_text(encoding="utf-8"))
session_key = payload.get("session_key", "")
data = payload.get("data", {})
text = data.get("text", "")
if not text or not session_key:
path.unlink(missing_ok=True)
continue
conn = sqlite3.connect(str(db_path))
try:
conn.execute(
"INSERT INTO messages (session_key, role, content, created_at) "
"VALUES (?, ?, ?, ?)",
(session_key, "user", text, payload.get("ts", int(time.time()))),
)
conn.commit()
recovered += 1
path.unlink(missing_ok=True)
except Exception as exc:
logger.warning(
"Failed to recover pending message for %s: %s",
session_key, exc,
)
# Re-save so next startup can retry
finally:
conn.close()
except Exception as exc:
logger.debug("Failed to read flush file %s: %s", path, exc)
path.unlink(missing_ok=True)
if recovered:
logger.info(
"Recovered %d pending message(s) from shutdown flush", recovered,
)
return recovered