Prometheus Certified Associate (PCA)
A multiple-choice, associate-level certification covering observability concepts, Prometheus architecture, PromQL, instrumentation and exporters, and alerting with Alertmanager.
- src
- Linux Foundation Training & Certification — Prometheus Certified Associate (PCA) certification page (training.linuxfoundation.org)
- chk
The PCA is a knowledge exam, and an easy one. Sixty multiple-choice questions in ninety minutes, and everything it tests is published free on prometheus.io. It does not prove you can operate Prometheus at scale — it proves you have read the documentation carefully. That is worth something, but be clear about which thing.
What it is actually good for
Learning PromQL properly. Most people who run Prometheus daily know four queries and copy the rest from dashboards someone else built. The exam forces you to learn rate versus irate, aggregation over time versus over dimensions, vector matching and histogram_quantile — and those are genuinely useful the day after you pass.
The badge itself is a weak signal. It is beginner-level, closed-book recall, with none of the CKA's live cluster. Treat the certificate as a side effect of the studying, not the reason for it.
Where the marks are
PromQL is 28% and the only domain you cannot pass by reading. Everything else is definitions. Weight your prep the same way: if you have twenty hours, spend ten of them writing queries against a running server.
The trap is the Observability Concepts domain at 18%. Logs, traces, spans, push versus pull and SLO/SLI/SLA basics are barely mentioned on prometheus.io, so candidates who study only the Prometheus docs walk in having skipped nearly a fifth of the exam. PromLabs, who helped define the exam scope, say the same thing about their own training.
Before you book
Closed book. No documentation, no Prometheus instance, no browser tabs. This is the opposite of the CKA, and it means memorising function names rather than knowing where to look them up.
75% is 45 of 60. Fifteen wrong answers is the whole margin, and multiple-select questions score all-or-nothing.
$250 with a free retake, valid two years. Renewal means sitting the current exam again before it expires — there is no maintenance fee and no continuing-education route.
New to Linux and the command line?
This path assumes fundamentals you may not have yet. Our Foundations Pack is out and free — Linux, the shell and Git, with exercises that mark your work and explain why you got it wrong. We're writing an agents pack next; leave your email if you want to hear when it ships.
One email when the pack launches. No spam, unsubscribe any time.
Exam domains
PromQL
28%Prometheus Fundamentals
20%Observability Concepts
18%Alerting & Dashboarding
18%Instrumentation and Exporters
16%Preparation path
- 1
Run Prometheus locally before reading about it
Download the binary, scrape itself and node_exporter, and open the expression browser. Two hours of a running server makes the architecture questions trivial later — targets, scrape intervals and the /metrics endpoint stop being abstractions once you have watched them work.
~6 hours - 2
Learn the data model, labels and exposition format exactly
Counter, gauge, histogram and summary, and what each one is for. This is 20% of the exam and it is where precise wording matters — questions distinguish a counter from a gauge, and a histogram from a summary, on details that are easy to half-know.
~10 hours - 3
Spend most of your study time on PromQL
The heaviest domain at 28%, and the only one you cannot learn by reading. Work through selectors, rate, aggregation over time versus over dimensions, binary operators and histogram_quantile — then write each one yourself against a live server until the syntax is automatic.
~25 hours - 4
Instrument one service yourself
Add a client library to a small application, expose counters and a histogram, and name them by the official conventions. Then read an exporter's source. Knowing why you instrument directly rather than writing an exporter is the distinction this 16% domain keeps testing.
~10 hours - 5
Write alerting rules and route them through Alertmanager
Configure a rule with a for clause, fire it, and watch Alertmanager group, inhibit and silence it. Grouping and inhibition are the parts candidates skip and the parts the exam asks about, because they are what separates Alertmanager from the rule evaluator.
~10 hours - 6
Close the observability-theory gap the Prometheus docs do not cover
Logs, traces and spans, push versus pull, and SLO/SLI/SLA definitions are 18% of the exam and largely absent from prometheus.io. Read them from a vendor-neutral source, because these are vocabulary questions and the published definitions are what gets marked.
~8 hoursObservability primer — logs, metrics, traces, spansOpenTelemetry · FreeService Level Objectives — Google SRE BookGoogle · FreeImplementing SLOs — SRE WorkbookGoogle · FreePCA exam preparation notesPromLabs · FreeObservability EngineeringCharity Majors, Liz Fong-Jones, George Miranda, Austin Parker · Paid · aff — Affiliate link — we may earn a commission at no extra cost to you. A free alternative is always listed alongside.
Frequently asked questions
Career Roadmaps
- Observability Engineer RoadmapA path into observability as a craft of its own — wide events, signal correlation, telemetry cost, collector pipelines, high-cardinality analysis, continuous profiling, and running observability as a platform other teams consume.
- Site Reliability Engineer RoadmapA path from DevOps fundamentals into the specialized discipline of site reliability engineering, covering SLOs, observability, incident response, data reliability, and capacity planning.