How to turn on logging on different threads properly?

[tisonkun]

Sometimes, when I enable logging in the main function like:

tracing_subscriber::fmt()
        .with_max_level(LevelFilter::INFO)
        .init();

.. and spawn threads like:

let rpc = ServerRpc::new(peer, tx_requests);
std::thread::spawn(move || rpc.do_main());

Logging functions in rpc.do_main are not captured. But if I change it to:

{
            // start server rpc
            trace_span!("server rpc");
            let rpc = ServerRpc::new(peer, tx_requests);
            std::thread::spawn(move || rpc.do_main());
        }

Logging functions in rpc.do_main work properly.

Is there some best practices I understand this case? How can I properly config tracing manner and how can I add context kv like slog Logger::new.

FYI - I don't know what trace_span does, just randomly place some statements as a guess.

[davidbarsky]

Logging functions in rpc.do_main are not captured. But if I change it to:

I'm not sure what you mean by this, but I'm guessing you're referring to the fact that logs in rpc.do_main don't have context.

Is there some best practices I understand this case? How can I properly config tracing manner and how can I add context kv like slog Logger::new.
FYI - I don't know what trace_span does, just randomly place some statements as a guess.

I would recommend reading the documentation on spans which explains how they are used and why changing the code as you did makes things work. spans do not have a full 1:1 analogue to slog's Logger, but they are roughly similar.

[tisonkun]

logs in rpc.do_main don't have context

No. It means the logs are not flush to the console, i.e., no output.