Path 05 / Learning syllabus
Observability
Answer operational questions using measurements with enough context and coverage to support the next decision.
Explore the syllabus- 01Start with a question, then find evidenceAvailable guide
- 02Follow the request and background workComing soon
- 03Read metrics without losing the populationComing soon
Your course project
An evidence-backed investigation
Investigate Dispatch exports using a synthetic evidence packet, then instrument a small prepared service. Build a request-oriented dashboard, inspect correlated logs and traces, test an instrumentation gap and document the resulting investigation.
Before you begin
Start with the foundations.
For engineers who can describe a basic service but are new to investigating it with telemetry. Lesson1 is a paper exercise; instrumentation lessons use a prepared local sample and link back to Linux fundamentals.
Each exercise states what you need and how to check its result. Use practice data and a disposable lab.
The ordered path
What you will learn
Follow the order, or open the guide that answers your question. Lessons marked Coming soon are not available yet.
- 01Available guide · 6 min read
Start with a question, then find evidence
Define the user symptom, time window and population before interpreting a dashboard.
What you will try
Use a supplied metric, log and trace to select the next check and reject an unsupported root-cause claim.
- 02Coming soon
Follow the request and background work
Map synchronous requests, queues and asynchronous completion before adding instrumentation.
What you will try
Identify what the status endpoint measures and which parts of an export happen elsewhere.
- 03Coming soon
Read metrics without losing the population
Understand counters, gauges, rates, distributions, labels and aggregation.
What you will try
Compare overall and tenant-specific slow-request fractions using defined windows and denominators.
- 04Coming soon
Write logs that explain an event
Preserve timestamps, event meaning and correlation fields without exposing sensitive payloads.
What you will try
Improve a sample log entry and explain which investigation question it can now answer.
- 05Coming soon
Read a distributed trace
Interpret spans, parent-child relationships, links, concurrency and unexplained time.
What you will try
Read a trace with overlapping work without summing child durations into a false request time.
- 06Coming soon
Instrument a small service with OpenTelemetry
Add a bounded set of useful signals and verify that they reach the intended destination.
What you will try
Run the prepared service and collector, create a known request and verify its recorded path.
- 07Coming soon
Connect signals across a request
Use consistent identifiers and context propagation to move from a symptom to related evidence.
What you will try
Follow one request across metrics, a log event and a trace, noting where correlation stops.
- 08Coming soon
Build a dashboard for an investigation
Choose panels and grouping that answer the service question instead of displaying every available metric.
What you will try
Build a view that exposes a tenant-specific slowdown hidden by an aggregate.
- 09Coming soon
Detect user impact and test the alert
Connect telemetry to an existing response objective and verify alert behavior.
What you will try
Exercise a supplied alert with a controlled failure and show what responders would actually receive.
- 10Coming soon
Manage gaps, sampling, cardinality and cost
Explain what collection decisions omit and how they affect an investigation.
What you will try
Compare full metric counts with sampled traces and choose a bounded label set and retention plan.
- 11Coming soon
Investigate a service failure with incomplete evidence
Form competing explanations, perform a discriminating check and retain unresolved uncertainty.
What you will try
Use the assembled evidence to investigate a staged failure and test whether the suspected cause fits.
- 12Coming soon
Review the evidence and its limits
Deliver an investigation another engineer can reproduce without overstating causality.
What you will try
Present the symptom, population, timeline, queries, decisive checks, missing evidence and next instrumentation change.
