discarding bad PubSub connection: i/o timeout... but I'm not using PubSub...

[wwarby]

I have relatively simple app which is using bloom filters. I'm creating my failover client like this, using a Redis sentinel environment:

redis.NewFailoverClient(&redis.FailoverOptions{
    MasterName:       "my-master",
    SentinelAddrs:    []string{ "localhost:26379" },
    ReadTimeout:      time.Millisecond * time.Duration(2000),
}

I'm making only two types of redis calls in the whole app: BF.INSERT ... and BF.MEXISTS ...

The app is spitting out the following messages to stdout at least once a second:

redis: 2022/08/18 09:34:49 pubsub.go:159: redis: discarding bad PubSub connection: read tcp 172.17.0.8:38868->10.101.245.231:26379: i/o timeout
redis: 2022/08/18 09:34:49 pubsub.go:159: redis: discarding bad PubSub connection: read tcp 172.17.0.8:38884->10.101.245.231:26379: i/o timeout
redis: 2022/08/18 09:34:49 pubsub.go:159: redis: discarding bad PubSub connection: read tcp 172.17.0.8:38940->10.101.245.231:26379: i/o timeout

This issue seems related and I can make the log messages go away by setting ReadTimeout on the client to -1, but that has an unwanted side effect (in that I actually want a read timeout for my bloom filter commands).

As far as I'm concerned my app is not using Redis PubSub functionality, so this behaviour must be originating from either the internals of Redis Bloom or have something to do with the fact that I'm using sentinels (I don't see it when I use a standalone single instance of Redis locally). Any idea what I can do to make these messages go away?

[VashingMachine]

I observe the same behavior in my application. It seems like the PubSub HealthCheck was trying to connect to the Redis Sentinel instead to Redis. I recreated this issue in here