refactor: 아키텍처 수정 — 동기HTTP→aiohttp + 연결 모니터링 + 재시도큐

#1 동기 HTTP → async aiohttp (Critical)
  - RemoteTransport: urllib.request → aiohttp.ClientSession
  - 모든 HTTP 요청이 non-blocking으로 전환
  - 이벤트 루프 블로킹 문제 해결

#2 연결 상태 모니터링
  - RemoteTransport: connected 플래그 + 연속 실패 카운터
  - Collector: 30초마다 health check → 실패 시 경고 로그
  - 연결 복구 시 ' Gateway connected' 메시지

#3 실패 재시도 큐
  - RemoteTransport: _retry_queue (최대 100건)
  - POST 실패 시 큐에 저장, 연결 복구 후 자동 재전송
  - Collector: 10초마다 retry flush
This commit is contained in:
Variet Worker
2026-03-11 22:55:54 +09:00
parent d2a477e12e
commit 1bf41ceee3
2 changed files with 153 additions and 71 deletions

View File

@@ -69,14 +69,17 @@ class CollectorBridge:
self._poll_commands_loop(),
self._forward_chat_snapshots_loop(),
self._forward_registrations_loop(),
self._health_check_loop(),
self._retry_flush_loop(),
]
if self.event_queue:
tasks.append(self._forward_events_loop())
await asyncio.gather(*tasks)
async def stop(self):
"""Stop the Collector."""
"""Stop the Collector and close HTTP session."""
self._running = False
await self.remote.close()
logger.info("[COLLECTOR] stopped")
# ─── Forward local pending → Gateway ───
@@ -114,7 +117,7 @@ class CollectorBridge:
continue # Skip pre-existing files from before startup
# Forward to Gateway (new or updated)
self.remote.write_json("pending", fname, data)
await self.remote.awrite_json("pending", fname, data)
self._forwarded_pending.add(rid)
self._pending_hashes[rid] = content_hash
@@ -143,7 +146,7 @@ class CollectorBridge:
try:
# Check each forwarded pending for a response
for rid in list(self._forwarded_pending):
data = self.remote.read_json("response", f"{rid}.json")
data = await self.remote.aread_json("response", f"{rid}.json")
if data is None or data.get("waiting"):
continue
@@ -166,7 +169,7 @@ class CollectorBridge:
"""Poll Gateway for commands and write them locally for Extension."""
while self._running:
try:
commands = self.remote.poll_commands(self.project_name)
commands = await self.remote.apoll_commands(self.project_name)
for cmd in commands:
cmd_id = cmd.get("id", str(int(time.time() * 1000)))
fname = f"{cmd_id}.json"
@@ -192,7 +195,7 @@ class CollectorBridge:
project = data.get("project_name", self.project_name)
content = data.get("content", "")
if content:
self.remote.send_chat(project, content)
await self.remote.asend_chat(project, content)
logger.info(f"[COLLECTOR] → Gateway: chat snapshot len={len(content)}")
f.unlink() # Cleanup after forwarding
except (json.JSONDecodeError, OSError) as e:
@@ -219,7 +222,7 @@ class CollectorBridge:
conv_id = data.get("conversation_id", "")
project = data.get("project_name", "")
if conv_id and project:
self.remote.register_session(conv_id, project)
await self.remote.aregister_session(conv_id, project)
forwarded_regs.add(f.name)
logger.info(f"[COLLECTOR] → Gateway: register {conv_id[:8]}{project}")
except (json.JSONDecodeError, OSError) as e:
@@ -246,9 +249,33 @@ class CollectorBridge:
"content": event.content,
"timestamp": event.timestamp,
}
self.remote._request("POST", "/api/event", event_data)
await self.remote.asend_event(event_data)
logger.info(f"[COLLECTOR] → Gateway: event {event.event_type.value} {event.file_name}")
except asyncio.TimeoutError:
continue
except Exception as e:
logger.error(f"[COLLECTOR] forward_event error: {e}")
# ─── Health check ───
async def _health_check_loop(self):
"""Periodically check Gateway connectivity."""
while self._running:
try:
ok = await self.remote.health_check()
if not ok and self.remote.connected:
logger.warning("[COLLECTOR] ❌ Gateway health check failed")
except Exception:
pass
await asyncio.sleep(30)
# ─── Retry flush ───
async def _retry_flush_loop(self):
"""Periodically flush failed request retry queue."""
while self._running:
try:
await self.remote.flush_retry_queue()
except Exception as e:
logger.error(f"[COLLECTOR] retry flush error: {e}")
await asyncio.sleep(10)