Observability
Observability is the ability to infer a system’s internal state and behavior from its externally available outputs, such as logs, metrics, and traces.
Effective Platform Engineering
Observability is the ability to determine and explain a software system’s internal state and usefulness from its output data.
- How it is used
- Uses output data to diagnose platform and product behavior and demonstrate value to engineering and business stakeholders.
- How deep it goes
- Defines observability, distinguishes it from monitoring, and outlines an incremental implementation sequence without specifying a complete deployment.
- Examples
- Uses failed scheduled tasks, delayed business reports, isolated dashboards, and Kubernetes outages to expose visibility gaps.
- What you can do after
- Build it from scratch incrementally by setting goals, instrumenting code, verifying collection, designing shared access, and assigning responsibility.
Read how Effective Platform Engineering explains observability →
Secure APIs
API observability is the ability to infer the health, behavior, and security-relevant activity of APIs from their telemetry.
- How it is used
- Uses telemetry to understand API behavior, detect threats, investigate incidents, and support timely response and recovery.
- How deep it goes
- Explains observability through logs, traces, metrics, instrumentation, threat signals, and sensitive-data filtering, while deferring detailed implementation beyond focused examples.
- Examples
- Uses endpoint error rates, request traces, malicious input, unauthorized responses, response latency, and filtered order-request logs as concrete examples.
- What you can do after
- Diagnose/debug with it by correlating telemetry, reproducing suspicious requests, identifying malicious activity, and excluding sensitive data from logs.
Grokking AI Applications
The capability to understand an application's internal behavior from its outputs, telemetry, logs, traces, and related runtime signals.
- How it is used
- Supports monitoring, debugging, and improving deployed GenAI applications by exposing agent behavior and orchestration steps.
- How deep it goes
- Explains step visibility through a LangSmith setup and trace, without descending into telemetry or framework internals.
- Examples
- Uses the HungryTraveler agent, MCP-connected itinerary and recipe tools, email delivery, and LangSmith traces.
- What you can do after
- Diagnose/debug with LangSmith by inspecting an agent's execution trace and tool-call sequence.
Cloud Observability in Action
Observability is the capability to infer and continuously discover a system’s internal state from externally available signals in order to influence that system.
- How it is used
- Guides teams from telemetry collection to actionable decisions about cloud-native systems, incidents, performance, costs, and user experience.
- How deep it goes
- Defines sources, signals, agents, destinations, and telemetry, then demonstrates correlation and operational trade-offs without detailing signal implementations.
- Examples
- Uses the Sock Shop Kubernetes application, shopping traffic, service logs, metrics, traces, dashboards, and role-based investigation flows.
- What you can do after
- Diagnose distributed-service problems, correlate signals, evaluate operational health, and influence systems through human or automated action.
Read how Cloud Observability in Action explains observability →
Architecting an Apache Iceberg Lakehouse
Observability is the capability to infer a system's internal behavior and resource usage from its emitted telemetry and operational data.
- How it is used
- Supports reliable ingestion, maintainable orchestration, catalog operations, and governed self-service analytics.
- How deep it goes
- Connects metrics, logs, metadata queries, dashboards, and alerts to operational decisions, without specifying complete monitoring configurations.
- Examples
- Uses ingestion metrics, Iceberg metadata queries, external dashboards, query logs, and compaction failure logs.
- What you can do after
- Diagnose failures and emerging performance or storage problems through correlated logs, metrics, metadata, and alerts.
Read how Architecting an Apache Iceberg Lakehouse explains observability →
Generative AI in Action
Observability is the practice of monitoring, logging, and tracing a system to understand its internal behavior and diagnose problems from its outputs and operational data.
- How it is used
- Supports production LLM application health checks, problem diagnosis, and decisions about scaling and optimization.
- How deep it goes
- Defines monitoring, logging, and tracing, then illustrates instrumentation with tools rather than developing a full production monitoring procedure.
- Examples
- Uses a timed chat-completion call, token counts, MLflow dashboards, and a Traceloop example.
- What you can do after
- Apply it via MLflow, Traceloop, or Prompt flow to monitor, log, and trace an LLM application.
Machine Learning Platform Engineering
Observability is the ability to infer the internal state and behavior of a deployed system from its outputs and telemetry, including application-specific information flows.
- How it is used
- Supports production monitoring, diagnosis, optimization, cost control, and review across conventional ML services and LLM applications.
- How deep it goes
- Traces infrastructure, model, retrieval, generation, metadata, errors, costs, and latency, stopping short of detailed implementation beyond Langfuse integration.
- Examples
- Uses ID-card detection, movie recommendation, DakkaBot RAG traces, Langfuse dashboards, and BentoML health and metrics instrumentation.
- What you can do after
- Diagnose with Langfuse by tracing RAG steps, retrieval results, model calls, failures, latency, token usage, and estimated API costs.
Read how Machine Learning Platform Engineering explains observability →
Automating API Delivery
Observability is the ability to query and analyze instrumented system behavior to determine the causes of previously unanticipated problems.
- How it is used
- Supports continuous production measurement by helping teams investigate unanticipated system problems.
- How deep it goes
- Defines observability and distinguishes it from predefined monitoring, without detailing instrumentation or tool setup.
- Examples
- Uses event data and logs as telemetry, alongside predefined metrics, dashboards, and alerts as monitoring artifacts.
- What you can do after
- Diagnose previously unanticipated production problems by querying and analyzing instrumented system behavior.
Troubleshooting Java, Second Edition
Observability is the ability to infer a system's internal behavior from its emitted telemetry and other external outputs.
- How it is used
- Uses telemetry to debug failures, monitor health, analyze performance, and follow requests across complex architectures.
- How deep it goes
- Explains instrumentation, export, collection, and visualization, then applies traces and percentiles to a timeout failure.
- Examples
- Uses a missing-order investigation and a frontend-to-order-service timeout mismatch with trace durations and latency percentiles.
- What you can do after
- Diagnose distributed failures by instrumenting services, inspecting traces, and comparing observed latency with configured timeouts.
Read how Troubleshooting Java, Second Edition explains observability →
AI Agents in Action, Second Edition
Observability is the practice of inferring and diagnosing a software system's internal state from its externally emitted telemetry.
- How it is used
- Supports debugging, incident resolution, evaluation, optimization, and production monitoring for agentic systems.
- How deep it goes
- Defines traces, metrics, and logs with implementation guidance, but stops short of detailed backend configuration.
- Examples
- Uses Phoenix dashboards, OpenTelemetry traces, correlation metadata, structured logs, evaluator checks, and deployment feedback systems.
- What you can do after
- Diagnose/debug with it by tracing interactions, correlating failures, measuring outcomes, and evaluating changes against curated tests.
Read how AI Agents in Action, Second Edition explains observability →
Logs and Telemetry
Observability is the ability to understand a system's internal state from externally emitted telemetry such as logs, metrics, and traces.
- How it is used
- Uses aggregated telemetry to diagnose failures, understand distributed behavior, and maintain operational awareness.
- How deep it goes
- Explains telemetry types, cross-system aggregation, and operational purpose without detailing instrumentation implementation.
- Examples
- Uses database storage failures, distributed applications, cross-service transactions, and automated build pipelines as operational examples.
- What you can do after
- Diagnose failures by preserving, aggregating, and correlating telemetry across distributed systems.
Digital Twins in Action
Observability is the ability to understand a system's internal condition and behavior from its externally generated outputs and operational signals.
- How it is used
- Uses observability to explain digital-twin discrepancies and decisions and to assess production readiness.
- How deep it goes
- Explains logging, metrics, and traces through a structured record, query, cross-service trace, and OpenTelemetry instrumentation listing.
- Examples
- Uses predicted-versus-measured temperature, a hallway motion-to-light trace, and a trace-identifier log query.
- What you can do after
- Apply it via OpenTelemetry to correlate operational signals and diagnose drift or latency in a digital twin.
Google Anthos in Action
Observability is the ability to understand a system's internal state from the outputs it produces, such as metrics, logs, and traces.
- How it is used
- Provides operational visibility across microservices, workloads, dependencies, and customer journeys.
- How deep it goes
- Explains proxy-based collection, monitoring destinations, dashboards, topology, and SLOs without detailing telemetry schemas or instrumentation internals.
- Examples
- Uses Anthos Service Mesh dashboards, topology maps, traffic-flow logs, service dependencies, and customer-journey SLOs.
- What you can do after
- Apply it via Anthos Service Mesh to inspect service relationships, review telemetry, and evaluate SLO and SLA impact without changing application code.
AI Agents in Action
Observability is the collection and analysis of execution signals that reveal an agent system's behavior, performance, and failures.
- How it is used
- Uses tracing and dashboards to inspect complex multi-agent behavior, performance, costs, outputs, and failures.
- How deep it goes
- Walks through installation, API-key configuration, initialization, execution, and dashboard inspection without developing tracing internals.
- Examples
- Uses a joke-generating crew and customer-review sentiment analysis to demonstrate dashboard monitoring and error detection.
- What you can do after
- Operate/configure AgentOps in CrewAI and evaluate agent performance, cost, outputs, inefficiencies, and interaction errors.
Bootstrapping Microservices, Second Edition: With Docker, Kubernetes, GitHub Actions, and Terraform
Observability is the practice of exposing and analyzing detailed telemetry to understand a system's internal behavior from its external outputs.
- How it is used
- Supports production monitoring, incident investigation, and reliability work across distributed microservices.
- How deep it goes
- Defines telemetry, querying, visualization, and correlation, but stops short of implementation code.
- Examples
- Uses Honeycomb traces and questions about users, regions, requests, messages, and error locations.
- What you can do after
- Diagnose/debug distributed failures by querying telemetry, correlating service interactions, and locating problematic microservices.
Building Distributed Applications that Work
Observability is the capability to infer the internal behavior and condition of a system from its externally emitted telemetry.
- How it is used
- Uses telemetry to verify distributed workflows and diagnose failures across service boundaries.
- How deep it goes
- Identifies logs, metrics, and distributed traces and explains their diagnostic role, but does not specify instrumentation steps here.
- Examples
- Follows a shopper’s request through a storefront, API, data stores, and background-processing components.
- What you can do after
- Diagnose/debug with it by following requests across services and identifying failures or performance bottlenecks.
Read how Building Distributed Applications that Work explains observability →
Related concepts
- trajectoryhas part · A trajectory is an ordered sequence of states, actions, events, or operations that records or describes progression through successive steps or over time.
- monitoringhas part · Monitoring is the ongoing observation and measurement of a system’s behavior, state, or operational signals to assess its condition and detect changes or problems.
- telemetryuses · Telemetry is automatically generated or collected operational data describing a system’s activity, state, behavior, or performance.
- OpenTelemetryimplemented by · OpenTelemetry is an open, vendor-neutral observability standard and framework for instrumenting software and generating, collecting, and exporting telemetry such as traces, metrics, and logs.
- performanceuses · Performance is the measurable effectiveness or operational behavior of a system under specified conditions, assessed through relevant qualities such as task results, speed, responsiveness, throughput, or resource use.
- loghas part · A log is a record of events, activity, or diagnostic information produced during the operation of a system or process, often ordered by time.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.