Hello! Could you explain please what is the point to have more than one replica of repository or dagit service? I've read about a possibility to have replicas but unfortunately can't find more details on it. Let's say I have three gRPC-server replicas with default RunLauncher. Does it means that pipeline launch can be distributed across replicas?

Hi Nazar - the way we recommend launching pipelines that can scale horizontally and be distributed across different machines is to use a different run launcher that launches the run in a more sophisticated orchestration environment (DockerRunLauncher, ECSRunLauncher, and K8sRunLauncher are all popular choices). We don't actually have great support for running replicas of the gRPC server currently (although it's something we would like to improve) - I'd be curious where you saw that.

Thank you, Daniel. I found it on the page https://docs.dagster.io/deployment/overview.

got it, thanks

Ok. Can I ask you one more question? If we have grcp server in docker container and it has a bind mount on repository directory. Do we have to always reload container or maybe we can use some api call to apply changes? For now it looks like without reloading we can se changes in code of existing repository entities, for example solids, but when we need to remove or add repository objects, for example schedules, then we need to reload grpc server.

generally if the underlying Python code has changed, the server needs to restart in order for those changes to get picked up

ok, got it, thanks