Vladimir Dementyev

Hey @d4rky-pl, Thanks for the new details. I've checked the diff and all the commits between 1.6.0 and 1.6.3, and the only suspicious change is this: https://github.com/anycable/anycable/commit/7eb49e88b484abc9fc40cf39325c14e3f8402ab3 tl;dr We now...

Thanks! Could you please give 1.6.6 a try? (And it would be great if you can turn debug logs on for a short period of time and share them).

> we can't run 1.6.1 because mrb images for 1.6.1 and 1.6.2 remain broken Ah, sorry. Pushed updated images: - 1.6.1: `anycable/anycable-go:d27d123-mrb` - 1.6.2: `anycable/anycable-go:8870917-mrb` Please, give 'em a try.

Hey @d4rky-pl, I'm back to the investigation. Glad we nailed it down to the Redis pub/sub component. Although can't say yet why is that happening. Could you please provide the...

Thanks! Actually, I think, we found it (kudos to @OuYangJinTing and #271): we were leaking Redis pub/sub subscriptions (they stayed active even if no clients are subscribed to them). I...

Here is the shared context from one of the projects using AnyCable GraphQL: ```ruby shared_context 'graphql:subscription' do let(:anycable) { AnyCable.broadcast_adapter } let(:fingerprint) { SecureRandom.uuid } let(:channel) do socket = double('Socket',...

Looks like when the `#clean` is called the `connection_class.connection` already points to a new connection, no the one that was used in the `#start`. Is there a test that reset...

@viralpraxis Let's add this issue on the TODO list for the end of May 😉

> each line should be robust and an exception in one line shouldn't affect the ability for the others to run Yeah, that's something I considered initially but decided to...

Hey @synth, I decided to implement a basic error handling feature. Now you can set `self.raise_on_line_error = false` in the base delivery class to rescue line delivery errors and report...