Remove re-subscription logic in scaling mode

[ashiquzzaman33]

This logic was originally introduced in #251 due to the use of createLazyClient(). Since it was replaced with createClient() in #281, it is no longer needed.

[joedixon]

Thanks @ashiquzzaman33. I understand it's no longer required when using createClient, but is there any reason not to have it in there in case an unsubscribe event is emitted? When using Redis for horizontal scaling, I can't think of a reason why you wouldn't want to ensure the subscription is established.

[ashiquzzaman33]

Thanks @joedixon ! 🙌
With createLazyClient() the unsubscribe event was synthetic — emitted automatically when the connection dropped — so auto-resubscribing was needed. With createClient() this isn’t the case: unsubscribe only fires when UNSUBSCRIBE (or closes the client) is explicitly called. Keeping the handler would override intentional unsubs and could even cause loops. For horizontal scaling, #281 already ensures reconnect + resubscribe, so this extra logic isn’t needed.

[ashiquzzaman33]

Also in RedisSubscribeClient.php , the subscribe() method doesn’t accept a channel parameter, so the current implementation of RedisPubSubProvider.php would actually throw an error when UNSUBSCRIBE called explicitly.

[joedixon]

The call to "subscribe" with the channel definitely looks like a bug so we'll get that fixed, but I still wonder about removing the "unsubscribe" listener. In the context of Reverb, when would you want to intentionally unsubscribe? I would argue we want to ensure the connection stays established at all costs.

[ashiquzzaman33]

You're right that we want to keep the connection established, and in the context of Reverb I can’t imagine a case where we’d intentionally unsubscribe. Since we never call unsubscribe, that event will never fire—connection loss triggers close, not unsubscribe, when using createClient. So this listener ends up as dead code. Horizontal scaling is already handled in #281, meaning keeping it adds no value and just leaves unused code.

[ashiquzzaman33]

Hi @joedixon !
Any update on this? Do you still want to keep this dead code?
If that’s the case, I can create a separate PR to fix below bug.

The call to "subscribe" with the channel definitely looks like a bug so we'll get that fixed

[joedixon]

I agree with you this code is no longer necessary. Thanks for the contribution.