16% of the CNPA exam. Sample questions below; the full library has 20 questions tagged to this domain.
Sample questions on Continuous Delivery & Platform Engineering
Continuous Delivery & Platform Engineering
Q1. Which of the following best describes Argo CD's reconciliation model?
Reveal answer and explanations
AArgo CD pushes manifests from the target cluster back to Git whenever it detects a drift event.
Incorrect. Argo CD reads desired state from Git and reconciles the cluster toward it; it does not push cluster state back to Git.
BArgo CD runs only once at install time and never reconciles further changes to the target cluster.
Incorrect. Argo CD reconciles continuously, not only at install time.
CArgo CD polls Git and the cluster and reconciles the cluster toward desired state declared in Git.
Correct. Argo CD continuously compares desired state in Git with actual cluster state and reconciles drift.
DArgo CD requires an external CI job to run `kubectl apply` on its behalf on every push to main.
Incorrect. Argo CD performs applies itself as part of reconciliation; no external CI step is required for this.
Continuous Delivery & Platform Engineering
Q2. A team uses Flux `HelmRelease` to deploy a chart, with `values` inlined and `valuesFrom` pointing at a `Secret`. On reconcile, Flux emits a `post-upgrade` hook failure. Which likely fact about Flux `HelmRelease` reconciliation applies?
Reveal answer and explanations
AFlux treats hook failures as a failed release and surfaces them on `HelmRelease.status.conditions`, with retries per policy.
Correct. Flux propagates helm hook failures to `HelmRelease.status`, and retry behavior depends on install/upgrade remediation settings.
BFlux ignores `post-upgrade` hook failures because they are purely informational and never affect the release state.
Incorrect. Hook failures are not informational; they affect release state.
CFlux automatically deletes the chart release when any hook fails, leaving no resources in the cluster whatsoever.
Incorrect. Automatic delete is not the default; remediation is policy-driven.
DFlux rewrites the values from the `Secret` into an in-line configmap; hooks never apply to rewritten values in any case.
Incorrect. Flux does not silently rewrite values in this manner.
Continuous Delivery & Platform Engineering
Q3. Which principle is central to a blameless postmortem?
Reveal answer and explanations
AIdentifying which individual is personally at fault and assigning a formal reprimand to that engineer.
Incorrect. Individual blame discourages transparency and reduces future learning.
BFocusing on systemic contributing factors and improvement opportunities rather than individual blame.
Correct. Blameless postmortems focus on systemic factors and improvements, not on punishing individuals.
CSkipping incident analysis when the incident did not cause significant customer-visible business impact.
Incorrect. Near-misses and low-impact incidents often yield the most valuable learning and should still be reviewed.
DKeeping the postmortem private to the on-call engineer to protect them from any reputational harm.
Incorrect. Sharing postmortems broadly is essential to organizational learning.
Continuous Delivery & Platform Engineering
Q4. Which of the four golden signals most directly surfaces service saturation for a bounded worker pool serving asynchronous tasks?
Reveal answer and explanations
ALatency of the first-response percentile measured for the primary external API endpoint used by customer clients.
Incorrect. Latency is a separate golden signal; it indicates symptom, not saturation of the worker pool.
BTraffic measured as requests per second routed through the public ingress controller for this particular service.
Incorrect. Traffic measures demand, not how full the worker pool is.
CNumber of 4xx HTTP errors returned by the service, aggregated weekly for executive review in a shared dashboard.
Incorrect. Error counts map to the errors signal, not saturation.
DQueue depth and in-flight worker count relative to the pool maximum, which directly indicates saturation pressure.
Correct. Saturation is measured by how close the system is to its capacity limit; queue depth and worker utilization capture this.
Continuous Delivery & Platform Engineering
Q5. Which of the following best represents a typical stage order in a CI pipeline?
Reveal answer and explanations
ASource checkout, build, unit test, static analysis, and artifact publish to a registry.
Correct. Source checkout leads to build, testing, analysis, and artifact publishing in a typical CI pipeline.
BDeploy to production, smoke test, build the binary, and finally check out the source code used.
Incorrect. Deploying to production before building the artifact is not a valid CI sequence.
CPerformance test first, then source checkout, then unit test, then a second source checkout step.
Incorrect. Running performance tests before source checkout and duplicating checkout stages is not typical.
DArtifact publish, then build, then checkout, then static analysis on the already-built artifacts.
Incorrect. Publishing an artifact before building or checking out is illogical.
Continuous Delivery & Platform Engineering accounts for 16% of the CNPA exam. Expect questions that test recall of terminology and the ability to read short scenarios — not deep configuration. Use the sample questions above as difficulty calibration; if any feel hard, the rest of our 20-question domain bank will close those gaps.