The bot builds a layered tracing_subscriber in main.rs. The layers are a
console formatter, an optional Tokio Console layer, and an optional
OpenTelemetry OTLP exporter — each with its own filter.
The console formatter's default filter is warn,serenity_discord_bot=info.
With RUST_LOG unset, the infra crates (serenity, h2, hyper, the gateway) are
quieted to warn so the bot's own info logs are visible. Setting RUST_LOG
overrides the default entirely.
Every instrumented span carries a custom category field for filtering. Values
include redis, ai_chat, llm, sql, and discord_command.
The top-level event_handler is deliberately not instrumented: it fires for
every gateway event, including presence updates, so a span there would be pure
noise. Only handled events carry spans.
Task-level async runtime inspection through
Tokio Console. It is feature-gated and
needs the tokio_unstable cfg at build time:
RUSTFLAGS="--cfg tokio_unstable" cargo run --features tokio_console
The OTLP layer is feature-gated behind opentelemetry and exports over
gRPC/tonic, so it can point at any OTLP-compatible collector. It uses a
separate filter from the console layer:
warn,serenity_discord_bot=info,tokio=off,runtime=off
tokio and runtime are turned off here so that, when tokio_unstable is
also enabled, the runtime-span firehose does not bury application traces in the
trace backend.
The compose stack ships Grafana Tempo and Grafana pre-wired as the UI. To bring up the telemetry backends without the bot:
docker-compose -f docker-compose.infra.yml up -dTo run Tempo manually, create /var/tempo once with your user as owner:
sudo mkdir -p /var/tempo && sudo chown $USER /var/tempo
tempo -config.file=./tempo.yaml
