Use tracing levels for different network events #53
Conversation
Everything was at `info`, which got pretty noisy. We now use `tracing` directly and leverage level to differentiate based on event frequency. Less frequent events (like new host or bind) use `info` and normal network traffic (like send/hold/recv etc.) use `trace`. Work still needs to be done here to integrate spans per #49.
| @@ -85,7 +85,7 @@ impl Host { | |||
| pub(crate) fn receive_from_network(&mut self, envelope: Envelope) -> Result<(), Protocol> { | |||
| let Envelope { src, dst, message } = envelope; | |||
|
|
|||
| trace!(?dst, ?src, protocol = %message, "Delivered"); | |||
| tracing::trace!(target: TRACING_TARGET, ?dst, ?src, protocol = %message, "Delivered"); | |||
There was a problem hiding this comment.
Isnt the target already set since the crate name is turmoil?
There was a problem hiding this comment.
It is the full mod path, which isn't major but it's pretty nice to just see turmoil on all the events.
There was a problem hiding this comment.
Interesting, @davidbarsky I wonder what you think
There was a problem hiding this comment.
Per #49 I assume everything that went to trace in this change would be a span. I don't like seeing the internal mod path in the traces, which is why I use a constant. It's not really a trace for the crate, rather it's a network trace that we happen to use tracing for. I don't feel too strongly though, just personal preference. Long term, I think we want a subscriber included that does nice things like write in different formats, and to different sinks like a file.
There was a problem hiding this comment.
sorry, i was offline yesterday for reasons (i still have a job thankfully, but I took a personal day).
Interesting, @davidbarsky I wonder what you think
@LucioFranco when i was talking to @mcches over discord, my feeling is that i'm eh about changing the target (overriding targets is a feature that is seldom used, and I expect that to remain the case until const evaluation develops some very advanced features, but that's out of scope for this discussion and it's better that turmoil expose/ship with a custom formatter or layer that does the things/formatting that Brett wants to do, but given the deadlines he mentioned, i feel that it's better to not let perfect become the enemy of good enough.
Everything was at
info, which got pretty noisy. We now usetracingdirectly and leverage level to differentiate based onevent frequency.
Less frequent events (like new host or bind) use
infoand normalnetwork traffic (like send/hold/recv etc.) use
trace.Work still needs to be done here to integrate spans per #49.