Patrick Sharp

What would be sending inline, though? Looking at the ActionCable source, it's calling publish in `Redis::Commands::Pubsub`, I believe. We're trying to eliminate possible causes so we're tracking any AC broadcast...

So looking at the backtrace, the line in actioncable that ends up at `call_v` is here https://github.com/rails/rails/blob/7-0-stable/actioncable/lib/action_cable/subscription_adapter/redis.rb#L92 and it's explicitly sending `_action_cable_internal`, which is sent all the way through to...

actioncable sets the message callback of the `Subscription` class to ```ruby on.message do |chan, message| broadcast(chan, message) end ``` If I'm reading it right, this block under `call_v` ```ruby while...

That is a concern, that since the exception being raised is new in this version, that it's something that could have been happening all along and was either being swallowed...

It does look like the gem uses RESP2 ```ruby def config(**kwargs) super(protocol: 2, **kwargs) end ``` The error itself says "inline request" and the comments on the method in the...

> Oh wait, this is an intermittent issue? Oh yes, intermittent and extremely elusive. We've had 0 luck in recreating it on purpose. We see the error 6-8 times a...

It looks like in the previous version, it looks like the errors were being raised in `call_loop` in client.rb, but the `ProtocolError` child class was overwriting `initialize` ```ruby # Raised...

Spent the time after my last attempt slow rolling the upgrade. Redis gem up to 4.8.1 and resque up to 2.5.0 with no issues. We have since moved from our...

hmm, in fact, I can't find us using ActionCable in the call that Airbrake captures as the source at all. Could be some sneaky metaprogramming somewhere, but I'm not finding...

@Matt-Yorkley, did you ever find any cause? I haven't been able to try again in a while, but I'm working on some more logging in the action cable process to...