How to timeout when Sentinel Master is unreachable

Is there a way to add a timeout to a get/set action? I’m looking for this in the docs and it’s either not there, or I’m somehow not seeing it.

I’m connecting ioredis to 3 sentinels using a master/slave setup. I can trigger a failover manually and everything works perfectly on the next request.

But, when I tell the current master to simulate an “unreachable” status: redis-cli -p 6379 DEBUG sleep 30

• the sentinels notice and elect a new master
• but the app waits for 30sec until the master is reachable again (now as a slave), then reconnects to the new sentinel master

My config is

{
    sentinels: [...the 3 sentinels],
    name: "mymaster",
    reconnectOnError: err => /^READONLY/.test(err.message),
}

[willrstern]

Also, here’s the Redis docs that show that sleep is a good way to simulate a master hanging. https://redis.io/topics/sentinel#testing-the-failover

[kapalkat]

So I have also followed described scenario and in my opinion this bug needs to be reopened. My setup: 3 sentinels, 3 redis slaves, 1 redis master. Sentinels env_vars: “REDIS_SENTINEL_DOWN_AFTER_MILLISECONDS”: “10000”. I am simulating Redis master down with command: redis-cli -p 6379 DEBUG sleep 60. Whats going on is: After 10 seconds, sentinels know right a way about the failure and elects new master. ioredis does NOT detect that! No error is passed. ioredis is still connected to the old master and my app hangs… After 60 seconds when its back, ioredis detects that old master is not a master anymore and tries to reconnect to the new one with a success. The question is: why ioredis does not get new master from sentinels?