OpenTelemetry Certified Associate (OTCA)¶
The vendor-neutral observability certification. OTCA covers instrumenting applications with the OpenTelemetry API and SDK, running the Collector, and reasoning about traces, metrics, and logs as a system.
This is the first observability certification in the repo. It is also, arguably, the most transferable one available: OpenTelemetry is the ingestion standard for Datadog, Grafana, Splunk, Honeycomb, New Relic, AWS, Azure, and Google Cloud. Learn OTel and you have learned the layer underneath every observability vendor.
Exam Details¶
- Exam Code: OTCA
- Duration: 90 minutes
- Questions: 60, multiple choice and multiple select
- Passing Score: 75%
- Cost: USD 250, includes one free retake
- Validity: 2 years
- Prerequisites: None
- Format: Knowledge-based, not hands-on
Full detail in the fact sheet.
Domains¶
| Domain | Weight | Notes |
|---|---|---|
| The OpenTelemetry API and SDK | 46% | 01 |
| The OpenTelemetry Collector | 26% | 02 |
| Fundamentals of Observability | 18% | 03 |
| Maintaining and Debugging Observability Pipelines | 10% | 04 |
The weighting is unusually lopsided. Nearly half the exam is instrumentation, and a quarter is the Collector. Together they are 72%, so that is where the study time goes.
What makes it tricky¶
OTCA is a knowledge exam like KCNA, not a hands-on exam like CKA. But the questions are specific: they ask which processor belongs in which pipeline position, what happens to context when a propagator is missing, and which instrument type fits a given measurement.
The three areas that catch people:
- API versus SDK. The API is what your code calls. The SDK is the implementation that actually records and exports. Without an SDK configured, API calls are no-ops. Library authors depend on the API only; applications wire up the SDK.
- Instrument selection. Counter, up-down counter, histogram, and gauge each fit specific measurement shapes, with synchronous and asynchronous variants. Questions describe a measurement and ask which instrument to use.
- Collector component ordering. Processors run in the order declared.
memory_limiterfirst,batchlast is the canonical guidance, and the exam tests whether you know why.
Study sequence¶
- Signals and the data model - traces, metrics, logs, and how they relate.
- API and SDK - instrumentation, context propagation, instruments, sampling.
- Collector - components, pipelines, deployment patterns.
- Fundamentals - cardinality, SLOs, golden signals, semantic conventions.
- Debugging - where telemetry gets lost.
Schedule in the practice plan.
Hands-on helps¶
Not required by the exam format, but the fastest way to make it stick:
- Run the Collector locally with an OTLP receiver and a debug exporter, and watch spans arrive
- Instrument a two-service application manually and confirm the trace joins across the boundary
- Break propagation deliberately by dropping the
traceparentheader and see two disconnected traces - Add an attribute with high cardinality to a metric and watch the series count explode
- Deploy the OpenTelemetry Operator and use auto-instrumentation injection
Study resources¶
- π OpenTelemetry documentation - the primary source; the exam tracks it closely
- π OTCA curriculum - published domains and competencies
- π Collector documentation - architecture and configuration
- π Semantic conventions - attribute naming standard
- Practice questions - question bank in this repo
Related¶
- PCA Prometheus Certified Associate - metrics depth
- KCNA - the cloud native fundamentals below this
- CKA - Kubernetes operations
- CNPA Platform Engineering Associate - the platform layer that consumes this
- Observability topic
- Set up a monitoring stack