work item listener does not recover from client disconnections
Closed this issue · 1 comments
the client work item listener does not have any retry logic to recover from errors when the stream closes or when the client looses the connection to the server
https://github.com/microsoft/durabletask-go/blob/main/client/worker_grpc.go#L36
the EOF error typically requires to create a brand new grpc connection, which is something that cant currently be done inside of the function StartWorkItemListener, additionally, it doesn't expose any means for users to detect that the streamed connection stopped, so users can handle the reconnection or retries...
I can't propose a solution, but IMO this definitely needs some refactor. I'm happy to contribute the improvements once a decision on how to address it is made
this has been implemented