Just started to work on gRPC project in golang and wonder how the context used in gRPC

Server

for example in scenario of streaming client on server side I have a function that implement the call:

rpc SendLogs(stream LogMessage) returns(google.protobuf.Empty) {}

func (s serverImpl) SendLogs(stream pb.xxxxServer) error {
}

First: what is this error that should be returned ? This is internal app (server) logic error ? Is this a status of .Recv() operation ?

Sometimes stream.Context().Err() already has an error of Deadline on Cancelled just before stream.Recv() – what does it mean ? Is this error propagated from client ? Should I return this error ?

Other RPC functions implementations (that does not stream – like unary call) received context as a first parameter – what is this context ? how it should be used ? Is it only to pass metadata between server and client ?

Client side

To create streaming client on client I first create gRPC server like this:

conn, err = grpc.NewClient(address, opts...)
...
...

client := pb.NewXXXXXXXClient(conn)
...

streamingClient, err = client.SendLogs(context.Background(), grpc.EmptyCallOption{})

what is this context parameter in in function ?

Could it be cancelled ?
In case canceling on this context make sense – what happens when it cancelled ? gRPC will close the stream and disconnect ? Receive GOAWAY response ?
Should I close conn.Close() and re-create gRPC client ? I understand that I can not continue and use context that was already cancelled ?

When I call streamingClient.CloseSend() function does it close the stream ? What will be received (?) on server side ?

Does gRPC will try to send multiple times begore giving up or I should take care of it ? Could I continue to send if error returned from .Send() function or the network stream is already unusable at this moment ?

How I could configure the gRPC itself to re-connect / re-try in case of network failures (such as time-out or ‘connection reset by peer’ error)