Troubleshooting
| Symptom | What to check |
|---|---|
| The broker does not start | docker compose logs roxa — config, license or port binding errors |
| A Kafka client cannot connect from outside | kafka_advertised_addr in serve.toml = an address visible to the client (not 127.0.0.1) |
| A Kafka client gets an authentication error | is SASL/PLAIN enabled? login/password from [[auth.users]]? code 58 = wrong credentials |
| No segments in storage | is the roxa bucket created? do the S3 keys in serve.toml match MinIO/.env? |
| Produce latency is growing | the Grafana dashboard: the S3 write lag (roxa_segment_put_latency_seconds) — a storage/network bottleneck |
| Metrics are empty | curl http://SERVER:9644/metrics | grep roxa_ |
| The web console does not see the broker | broker in admin.toml; broker_user/password = [[auth.users]] |
Logs
bash
docker compose logs -f roxa # the broker
docker compose logs -f roxa-admin # the web consoleThe logging level: the log field in the config or the RUST_LOG environment variable (for example, RUST_LOG=debug).
Metrics
The broker's Prometheus endpoint: :9644/metrics. Key series: roxa_produce_latency_seconds, roxa_fetch_latency_seconds, roxa_segment_put_latency_seconds (+ request counters). In the production stack, Grafana with the “Roxa — Overview” dashboard is provisioned automatically (:3000).
Did not find the answer — write to support: your question goes straight to the team.