Logs

Once telemetry is started, log lines written with the logging package, or any other use of Erlang's logger, join the trace they were written in:

• Warnings and worse become log events on the current span, with the level and the message, so they show up in your tracing backend next to the work that caused them.
• A process that crashes outside any span becomes a failed process crash span, so a crash in a background process is not lost.
• With a recorder, every line is kept for the admin's *Logs* page, linked to its trace.

JSON logs

json_logs writes each line as one JSON object with the trace and span ids, for log collectors that link logs to traces, such as Loki or CloudWatch:

logging.configure()
let assert Ok(Nil) =
  telemetry.new("acme-web")
  |> telemetry.otlp
  |> telemetry.json_logs
  |> telemetry.start

{"time":"2026-09-25T12:06:47.045Z","level":"warning","message":"slow notes","trace_id":"cfe5d76ab9bf1510ca60b9449f47eb37","span_id":"b7ad6b7169203331","source":"notes:list/1"}

Call start after logging.configure(), which sets a formatter of its own.

Log lines go wherever your logs go already. Anything a line holds, such as an address in an error message, goes with it, to your log collector and, for warnings and worse, to your tracing backend too.