Skip to content

Redis Connection Lost

The Redis adapter loses its connection mid-operation. Subsequent operations raise AdapterConnectionError.

Symptoms

  • Operations that use Redis (caching, pub/sub, queue backing) start raising AdapterConnectionError
  • ping() returns False
  • OTel spans show AdapterConnectionError on cache.* or queue.* child spans

Diagnosis

1. Verify Redis is reachable.

redis-cli -u <REDIS_URL> ping
# → PONG if reachable

2. Check if the connection was lost due to a timeout.

Redis servers close idle connections after timeout seconds (default: 0 = never, but managed Redis services often set this to 300s). If the adapter holds a single connection and doesn't reconnect, it fails after the server closes the idle connection.

Recovery

redis-py with auto_reconnect=True (the default) reconnects automatically on the next operation. If the adapter is not reconnecting:

  • Verify the adapter uses redis.asyncio.Redis with retry_on_timeout=True
  • TracingProxy is safe for reconnects — it resolves the method fresh on every call, so the new connection's methods are called correctly after reconnect

If Redis itself is down: restore Redis first. The adapter will reconnect automatically when Redis returns.

Prevention

Use connection pooling (ConnectionPool) rather than a single connection. Pool connections are acquired per operation and returned immediately, eliminating idle timeout issues.