Module 01 · 3 lessons
Evidence foundations
Define trustworthy telemetry identity, quality, and Spring instrumentation before building operational views.
17 reading lessons · 1 deterministic capstone
Operate the evidence chain, not just the dashboard.
Follow one Spring order service from instrumentation and telemetry quality through JVM signals, SLO alerts, incident command, and verified recovery.
Five ordered modules · evidence first
Move from trustworthy signals to verified recovery.
Each lesson separates emitted telemetry, observed evidence, user impact, and durable business outcomes.
Module 02 · 9 lessons
Production signals
Use bounded logs, metrics, traces, JVM recordings, GC, thread, and JIT evidence according to each signal's boundary.
- Structured Logging and Correlation IDs Explained
- RED, USE, and Golden Signals Explained
- Distributed Tracing and Context Propagation Explained
- JVM Profiling and JFR for Production Backends
- JVM Garbage Collection Logs and Pause Diagnosis Explained
- G1 Garbage Collector Diagnosis and Tuning
- ZGC Low-Latency Garbage Collection Diagnosis and Tuning
- JVM Thread Dumps, Safepoints, and CPU Saturation Diagnosis
- JVM JIT, Code Cache, and Warmup Diagnosis
Module 03 · 3 lessons
Telemetry delivery and reliability objectives
Operate the telemetry chain, define user-centered objectives, and route actionable alerts to owned runbooks.
Module 04 · 2 lessons
Incident operations
Coordinate response, communicate facts, learn without blame, and verify recovery with independent evidence.
Module 05 · Capstone lab
Capstone lab
Compare modeled user impact with observed evidence, SLO burn, alert windows, telemetry gaps, and recovery.
Browser-only capstone
Compare user impact with the evidence operators can actually see.
The lab is not a benchmark, capacity calculator, SLO compliance audit, incident replay, alerting dry-run, MTTA/MTTR calculator, or production verification.
Course library
Review all readings and the capstone guide.
Upgraded lessons keep their canonical slugs and browser-local progress keys.
System Design Aug 2, 2026 12 min read
Observability for Backend Systems Explained
Learn how logs, metrics, traces, and events work together to diagnose backend failures, protect telemetry quality, and verify recovery.
System Design Aug 19, 2026 4 min read
Telemetry Quality, Semantic Conventions, and Cardinality Explained
Build trustworthy telemetry contracts for identity, schemas, units, timestamps, missing data, duplicates, and bounded dimensions.
Backend Aug 19, 2026 4 min read
Spring Boot, Micrometer, and OpenTelemetry Instrumentation
Instrument a Spring order service with Actuator, Micrometer Observation, tracing, OTLP, bounded dimensions, and tested propagation.
System Design Aug 2, 2026 11 min read
Structured Logging and Correlation IDs Explained
Design searchable JSON logs and correlation IDs that connect requests, messages, traces, and failures without leaking secrets or exploding cost.
System Design Aug 2, 2026 11 min read
RED, USE, and Golden Signals Explained
Use RED, USE, and the four golden signals to diagnose request and resource failures with PromQL, safe labels, histograms, and exemplars.
System Design Aug 2, 2026 10 min read
Distributed Tracing and Context Propagation Explained
Trace backend work across HTTP and message queues with W3C context, safe baggage, meaningful spans, sampling boundaries, and broken-trace tests.
Java Aug 19, 2026 4 min read
JVM Profiling and JFR for Production Backends
Use Java 25 Flight Recorder to investigate CPU, allocation, GC, locks, I/O, and virtual threads without overstating profile evidence.
Java Aug 19, 2026 5 min read
JVM Garbage Collection Logs and Pause Diagnosis Explained
Read Java 25 garbage collection logs to distinguish allocation pressure, live-set growth, concurrent work, pauses, and Full GC safely.
Java Aug 19, 2026 5 min read
G1 Garbage Collector Diagnosis and Tuning
Diagnose Java 25 G1 young, marking, mixed, evacuation, and humongous-object behavior before making minimal evidence-backed tuning changes.
Java Aug 19, 2026 5 min read
ZGC Low-Latency Garbage Collection Diagnosis and Tuning
Diagnose generational ZGC allocation headroom, concurrent work, stalls, heap sizing, and latency-throughput tradeoffs on Java 25.
Java Aug 19, 2026 4 min read
JVM Thread Dumps, Safepoints, and CPU Saturation Diagnosis
Diagnose Java 25 platform and virtual threads, locks, safepoints, CPU saturation, pinning, and dependency waits with bounded evidence.
Java Aug 19, 2026 5 min read
JVM JIT, Code Cache, and Warmup Diagnosis
Diagnose Java 25 tiered compilation, compilation queues, deoptimization, code cache pressure, warmup, JFR, and AOT profile boundaries.
System Design Aug 2, 2026 11 min read
OpenTelemetry Collector Pipeline Explained
Build a resilient vendor-neutral telemetry pipeline with OTLP receivers, memory limits, batching, redaction, retries, queues, and self-monitoring.
System Design Aug 2, 2026 11 min read
SLI, SLO, Error Budgets, and Burn-Rate Alerts Explained
Turn user journeys into measurable SLIs, realistic SLOs, error budgets, and multi-window burn-rate alerts that page on sustained impact.
Operations Aug 19, 2026 4 min read
Production Alerting, On-Call, and Runbooks
Route symptom-based SLO alerts through actionable pages, tickets, ownership, escalation, handoff, and recovery-focused runbooks.
Operations Aug 19, 2026 4 min read
Incident Command, Communication, and Postmortems
Coordinate incident command, operations, communication, handoffs, factual timelines, blameless learning, and owned follow-up work.
System Design Aug 2, 2026 11 min read
Production Incident Troubleshooting with Logs, Metrics, and Traces
Follow an evidence-driven incident runbook from SLO alert and triage through metrics, traces, logs, mitigation, recovery verification, and postmortem.
Reliability Labs Aug 19, 2026 4 min read
SLO Incident Triage Lab: Evidence Gaps, Alerts, and Recovery
Run six deterministic browser scenarios that compare modeled user impact with observed telemetry, alert windows, and recovery evidence.
FAQ
Production Observability course questions
Does this course connect to production telemetry?
No. Every command and query is explanatory text, and the capstone is a deterministic browser-only model with no network, credentials, storage, telemetry query, or analytics write.
Does a cleared alert prove recovery?
No. Recovery requires fresh user outcomes, trustworthy evidence delivery, dependency capacity, and durable order-state verification.
Are retained traces a complete request history?
No. Sampling, redaction, pipeline loss, backend retention, and instrumentation gaps define what a trace set can support.
Can the capstone calculate a production SLO or MTTR?
No. Requests, steps, coverage, burn, and recovery are explicit teaching units rather than measurements, compliance results, forecasts, MTTA, or MTTR.