Advanced
Observability
Logs, metrics, traces — and the questions they answer.
Structured logging with correlation, the ELK stack, Prometheus metrics through Micrometer, Grafana dashboards and alerts that mean something, and distributed tracing with OpenTelemetry across services and Kafka.
5 lessons written1 modules~1h reading
After this course you can
- Emit structured logs with correlation ids and query them
- Define RED and USE metrics for a service and alert on symptoms, not causes
- Trace a request across three services and a Kafka topic
Curriculum
5 lessons · outlined lessons show their plan01
Structured loggingLevels that mean something, a disabled DEBUG line that still cost hundreds of nanoseconds, grep matching the wrong order where a JSON field did not, and the correlation id a thread pool loses.13 minWhat never to logThe list, and the reasonable code that breaks it: a generated toString, a record, an exception quoting its input, a header dump, a connection URL — each shown leaking, then fixed.11 minMetrics with Micrometer and PrometheusGolden signals, a mean of 45.6 ms hiding a p99 of 1.7 s, two p99s averaged to 860 ms when the truth was 49, histogram buckets as estimates, and cardinality.14 minDistributed tracingSpans and the traceparent header, a waterfall that names the slow SELECT, the thread pool that split a trace in two, and head sampling that kept 13 of 1,003 failures.13 minAlerting and SLOsSLIs, SLOs and the error budget, burn rate across windows, and a simulated month: a static threshold paged 20 times, 11 for nothing; burn-rate alerts paged 3, once per incident.14 min