RecursionError in redis.asyncio when using retry_on_timeout or health_check_interval

[everpoint-saas]

This might be related to #2893 (Sentinel recursion), but it also happens in non-Sentinel setups

Bug Description
redis.asyncio enters infinite recursion when retry_on_timeout=True or health_check_interval>0 are set.

Environment

Python: 3.11.x, 3.12.x (both affected)

redis-py: 5.0.1 (also tested with 4.6.0)

Parser: default (no hiredis)

OS: Docker (python:3.12-slim-bookworm)

uvloop: disabled

Minimal Reproduction

import asyncio
import redis.asyncio as redis

async def reproduce():
client = redis.from_url(
"redis://default:password@localhost:6379/0",
retry_on_timeout=True,
health_check_interval=30,
)
await client.ping() # RecursionError

asyncio.run(reproduce())

Working Configuration

client = redis.from_url(
"redis://default:password@localhost:6379/0",
retry_on_timeout=False,
health_check_interval=0,
)

Stack Trace Pattern
RecursionError: maximum recursion depth exceeded

connect() → on_connect() → send_command("CLIENT SETINFO")

→ check_health() → _send_ping() → send_command("PING")

→ back to connect() (infinite loop)

Key Findings

Trigger: enabling retry or health check during initial connection.

Not Python-version-specific, reproducible on 3.11 and 3.12.

Not parser-specific (occurs with PythonParser too).

Root cause: circular dependency between connection setup and health check.

Workaround
Disable retry and health check:

redis.from_url(..., retry_on_timeout=False, health_check_interval=0)

Expected Behavior
Health checks and retries should not trigger recursion during connection initialization.

[petyaslavova]

Hi @everpoint-saas, thanks for letting us know about this! We'll have a look at it soon.