Google Professional Cloud Developer Practice Test

Answer choices

1. A. Verify every API error is retried forever with exponential backoff so the client eventually succeeds without classifying errors as retryable or permanent.
2. B. Verify non-transient or non-retryable errors are not retried indefinitely.
3. C. Verify retries ignore server throttling.
4. D. Verify successful requests are duplicated.

Wrong-answer review

• A. Verify every API error is retried forever with exponential backoff so the client eventually succeeds without classifying errors as retryable or permanent.: Incorrect. Retrying every error forever can loop on permanent failures such as invalid arguments or authorization errors, consuming resources and potentially duplicating side effects. Retry policy must distinguish transient failures.
• C. Verify retries ignore server throttling.: Incorrect. Ignoring server throttling causes retries to compete with a service that is explicitly asking clients to slow down. Retry tests should verify appropriate backoff and respect for retryable conditions.
• D. Verify successful requests are duplicated.: Incorrect. Successful requests should not be deliberately duplicated by a retry library. Retrying after success risks duplicate mutations and violates normal retry semantics.

Answer choices

1. A. Embed the service-account key in the container image.
2. B. Place the service-account key in a public Cloud Storage bucket under an obscure object name so the container can download it at startup.
3. C. Use Application Default Credentials with the Cloud Run service identity.
4. D. Grant every developer Owner so credentials are unnecessary.

Answer choices

1. A. Supply the value as a Cloud Run environment variable or service configuration rather than hard-coding it in the image.
2. B. Store the nonsecret value in Secret Manager and grant the service secret-access permissions in every environment.
3. C. Bake the environment-specific value into the container image and publish a different image for each environment.
4. D. Pass the value as a Docker build argument and set it permanently in the image during the build.

Answer choices

1. A. Use `kubectl explain` from Cloud Shell to inspect resource fields while editing the manifest separately.
2. B. Use Cloud Code's Kubernetes development support in the IDE.
3. C. Edit the manifest in the Google Cloud console and validate it only when applying it to the cluster.
4. D. Use a generic YAML linter in the IDE without Kubernetes resource schemas or cluster-aware tooling.

Answer choices

1. A. Set `imagePullPolicy: Always` so each pod fetches the latest image before it starts.
2. B. Enforce a Binary Authorization policy that requires the approved attestation.
3. C. Grant Artifact Registry read access only to the production node identities and treat repository IAM as the signature check.
4. D. Use a custom admission webhook that checks image tags but does not verify trusted attestations for the digest.

Answer choices

1. A. Retry only retryable transient failures using bounded exponential backoff, adding jitter where appropriate.
2. B. Retry `UNAVAILABLE` with a fixed one-second delay forever until the call succeeds.
3. C. Apply the same retry loop to transient errors and permanent errors such as `PERMISSION_DENIED`.
4. D. Disable retries entirely so the client never risks issuing the operation twice.

Answer choices

1. A. Promote digest X rather than rebuilding an untested replacement artifact.
2. B. Rebuild the same source again in the production pipeline and deploy whatever new digest results after production dependencies are resolved.
3. C. Use the latest tag regardless of digest.
4. D. Choose the larger image because it contains more packages.

Answer choices

1. A. Log every request twice to compensate.
2. B. Tune trace sampling to retain a useful representative subset that meets troubleshooting needs.
3. C. Disable all telemetry.
4. D. Increase sampling to 100% solely to reduce cost.

Answer choices

1. A. Store all PDF bytes in every order row by default.
2. B. Store transactional rows as Secret Manager versions.
3. C. Keep transactional order metadata in a database and store invoice PDFs in Cloud Storage, linking them by identifier.
4. D. Use Pub/Sub as permanent PDF storage.

Answer choices

1. A. Cache the hot reference data in Memorystore with an expiration strategy appropriate to the update frequency.
2. B. Increase database write capacity.
3. C. Persist the cache only in each ephemeral application instance with an expiration policy chosen for the application's current data-refresh pattern.
4. D. Replace asynchronous requests with synchronous calls.