Redis lost connection does not seem to get re-created

[Sh4d1]

Hello, when running sequin in k8s, it seems every time my redis restarts, sequin gets blocked I and I get these errors :

Last message: {:phoenix, :async_result, {:start, {#Reference<0.0.3396227.629622195.1249705985.119942>, nil, :consumers_task, {:exit, {{:badmatch, {:error, %Sequin.Error.ServiceError{code: "no_connection", message: "[redis]: No connection to Redis", service: :redis, details: nil}}}, [{SequinWeb.SinkConsumersLive.Index, :"-load_consumer_metrics/1-fun-0-", 1, [file: ~c"lib/sequin_web/live/sink_consumers/index.ex", line: 168]}, {Enum, :"-map/2-lists^map/1-1-", 2, [file: ~c"lib/enum.ex", line: 1714]}, {Map, :new_from_enum, 2, [file: ~c"lib/map.ex", line: 262]}, {SequinWeb.SinkConsumersLive.Index, :load_consumers, 3, [file: ~c"lib/sequin_web/live/sink_consumers/index.ex", line: 147]}, {Phoenix.LiveView.Async, :do_async, 5, [file: ~c"lib/phoenix_live_view/async.ex", line: 220]}, {Task.Supervised, :invoke_mfa, 2, [file: ~c"lib/task/supervised.ex", line: 101]}]}}}}} line=2646 pid=<0.26533.0> file=gen_server.erl domain=otp
[error] [08:25:01.621] Redis error while trying to acquire mutex: {:error, %Sequin.Error.ServiceError{code: "no_connection", message: "[redis]: No connection to Redis", service: :redis, details: nil}} line=25 pid=<0.20162.0> file=lib/sequin/mutex.ex domain=elixir application=sequin mutex_key=sequin:mutexed_supervisor:prod:Elixir.Sequin.Runtime.MutexedSupervisor

It gets unblocked when I restart the sequin pod. I'm using sequin v0.13.14

[dosubot]

Sequin does not currently have explicit logic to automatically reconnect to Redis after a Redis server restart in Kubernetes. The connection cache will recreate connections if the process is dead, but if the process is still alive while the underlying socket is broken (which can happen after a Redis restart), Sequin won't detect this and won't recreate the connection. This is why the pod restart is required to recover. There are no recent changes or configuration options that address this in newer versions either, and this behavior matches the current code and issue history (source, source).

To work around this, you could use a Kubernetes liveness probe that triggers a pod restart if Sequin can't reach Redis, or implement external monitoring to restart the pod when Redis is unavailable. There is no built-in solution in Sequin for automatic reconnection after Redis restarts.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}    

[Sh4d1]

@RTLS is this expected ? 🤔