Hacker News Reader

Centralized logging is great, but recently one of my coworkers updated our rust backend code to use tracing and opentelemetry, and I've found it to be a pretty nice upgrade over regular logging. However it's more complicated to set up and isn't as widely supported as logs (which on most platforms is just a matter of logging to stdout).

If you had an app running on fly and wanted to use tracing, how would you set that up today?

I believe that logs should have the rule of three applied to them. The third time you use the logs to identify an issue, that should be converted to a stat.

Ultimately if the logs get wordier over time, the ability to spot an issue goes down, and some log messages are very expensive. If it’s not actionable, why are you logging it?

Glad to see it's NATS and not Kafka for the pubsub system.

... because?

I was just about to start standardizing logging on one of my apps deployed with Fly.io so perfect timing. The monitoring tool is OK, but pushing it to a proper logging tool that I can query will be much more helpful in the future.

Shipping Logs