Logging & Monitoring for Production

09:00 – 10:15

Turn a log file into something a machine can query

structured JSON logs into Loki, log levels used deliberately, what never gets logged

Takeaway An unstructured log converted and then queried for a condition that was previously ungreppable.

10:30 – 11:45

Follow one request across two services

correlation IDs, MDC and thread context, logging once per request

Takeaway Every line for one request retrieved in a single query, across a service boundary.

12:00 – 13:00

Instrument metrics that answer a question

Prometheus counters and gauges, histograms and cardinality, the four golden signals

Takeaway A latency histogram with p95 read off it, and a high-cardinality label shown exploding the store.

14:00 – 15:15

Build the one dashboard people actually use

rate errors duration in Grafana, baselines, annotating deploys

Takeaway A dashboard that answers "is it up, is it slow, is it failing" in under ten seconds.

15:30 – 17:00

Trace a slow request across three services

OpenTelemetry spans, sampling, reading a flame graph

Takeaway The responsible span identified in under a minute, end to end.

09:00 – 10:15

Write an alert worth waking up for

symptoms vs causes, thresholds and flapping, runbooks

Takeaway One alert written, fired on purpose, and resolved by following its own runbook.

10:30 – 11:45

Deploy without dropping a request

liveness vs readiness, graceful shutdown, draining

Takeaway A restart under load with no failed request, proved by the load generator, not asserted.

12:00 – 13:00

Keep the pipeline legal and affordable

personal data in logs, retention, cost drivers

Takeaway A pipeline checked for personal data, and a retention setting you can defend.

14:00 – 15:15

Diagnose a live incident from evidence alone

symptom to cause, discarding hypotheses

Takeaway An unfamiliar failure worked from logs, metrics and traces to a named cause.

15:30 – 17:00

Write the timeline and close the loop

incident timelines, what instrumentation was missing

Takeaway A written diagnosis, and the list of what you would instrument before the next incident.