1.1 Current Exam, Sources, and Scenario Method
Key Takeaways
Google lists 50-60 multiple-choice and multiple-select questions, two hours, and a USD 200 fee plus applicable tax.
The current blueprint weights are approximately 32% design, 23% build/test, 24% deployment, and 21% integration.
There are no prerequisites; the recommended experience is three years in industry including one year designing/managing Google Cloud solutions.
The current professional retake limit is four attempts in two years, with waits of 14, 60 and 365 days after successive failures.
Use official registration and policy pages for delivery, accommodations and renewal; no numeric passing threshold is published.
Current Exam and Independent Study Method
The Google Cloud Professional Cloud Developer credential concerns developers who design, build, test, configure and integrate cloud-native applications. It is distinct from Associate Cloud Engineer, which emphasizes operating cloud environments, and Professional Cloud DevOps Engineer, which emphasizes delivery and reliability engineering. There is overlap in tools, but the developer exam repeatedly asks what application code, configuration and service selection should do for a stated requirement. Read the requirement before recognizing a familiar product name.
This is independent study material by OpenExamPrep. Official exam information comes from Google, and product documentation controls the behavior of the individual services. A guide can teach a pattern without being an official endorsement or a substitute for hands-on experience. Check the current official certification page and its linked PDF rather than relying on a cached older web outline: older outlines have different domain percentages and topic lists.
Verified Logistics and Policy Boundaries
| Item | Official information checked October 18, 2026 |
|---|---|
| Format | 50-60 multiple-choice and multiple-select questions |
| Duration | Two hours |
| Registration fee | USD 200 plus applicable tax |
| Languages | English and Japanese |
| Prerequisites | None |
| Recommended experience | Three or more years of industry experience, including one or more years designing and managing Google Cloud solutions |
| Delivery | Pearson test centers or online proctoring, scheduled through Google's candidate portal |
| Numeric passing threshold | Not published; results are pass/fail |
| Professional certification validity | Two years |
The recommended experience is not an eligibility requirement. A learner can register without a prerequisite certification, but memorizing product names does not replace the experience needed to reason about failures and trade-offs. Likewise, a pass/fail result cannot be translated into a guaranteed number of correct answers. Do not infer an official pass percentage from a third-party practice test.
Google's current retake policy allows up to four attempts in a two-year period for associate/professional exams. After the first failure, wait at least 14 days; after the second, 60 days; after the third, 365 days. Attempts in different languages or delivery methods still count. Each attempt requires payment. Use the portal for current scheduling, accommodations, identification and cancellation procedures. The February 2026 transition ended Kryterion delivery; old Webassessor instructions are historical. Do not assume a language extension or accommodation is automatically approved.
Use the Blueprint to Allocate Work
The current domains are designing highly scalable, secure and reliable cloud-native applications (~32%), building and testing (~23%), configuring cloud-native applications for deployment (~24%), and integrating with Google Cloud services (~21%). The percentages are approximate blueprint weights, not a promise of a particular count of scored questions in one appointment.
A useful study cycle begins with a diagnostic against the objectives. Mark an objective as ready only when you can explain its mechanism, configure a small example, and identify a failure mode. For example, knowing that ADC exists is different from knowing its search order, why a local ADC file can shadow an attached identity, and why an authenticated caller can still lack an IAM permission. Keep the gap specific enough to practice.
For a hypothetical 100-hour review budget, the domain weights suggest an initial 32/23/24/21-hour allocation. Then adjust based on demonstrated gaps. This is an independent planning example, not Google's recommended preparation duration. An experienced backend developer may need more GKE configuration practice; an infrastructure engineer may need more transactions, SDK lifecycle and test isolation. Reassess with new scenarios rather than repeating only questions already memorized.
A Repeatable Scenario Method
- Extract the hard constraints: latency, durability, freshness, geography, credentials, protocol, operational effort and cost.
- Identify the boundary where the failure occurs: application, identity, network, platform lifecycle, database or API quota.
- Select a mechanism that changes that boundary. A larger IAM role does not create a private network route; more memory does not fix an invalid token audience.
- Check correctness under retries and scale. An asynchronous effect must survive duplicate delivery, worker failure and acknowledgment loss.
- Compare operational trade-offs. A fully managed service often reduces management work, but it does not automatically satisfy every protocol or state requirement.
Suppose a Cloud Run service returns success before a payment is durably recorded. Enabling always-allocated CPU can address between-request throttling, but cannot guarantee completion when the instance is terminated. The stronger design uses a durable queue or transaction/outbox, an idempotent worker and a success acknowledgment after durable work. This separates a narrow runtime symptom from the business requirement.
Suppose a private database connection times out. Check private routing, DNS, firewall/egress and connector configuration before changing SQL grants. If the connection reaches the database but login fails, inspect identity and database authentication. If login succeeds but a statement is rejected, inspect SQL privileges. The same word, permission, can hide several independent control planes.
Practice and Review Without Creating New Risks
Use a separate test project, synthetic data and small resource limits. Record the commands, expected result and cleanup procedure for each lab. Inspect generated configuration instead of assuming the console selected the intended default. For anything destructive, use disposable resources and verify the selected project. Review roles assigned to the build identity separately from the runtime identity.
After a quiz, explain why the selected mechanism satisfies every hard constraint and why each distractor misses one. A service can be useful in general and still be wrong for this scenario. Track misconceptions such as confusing request rate with concurrency, retention with backup, or authentication with network ingress. These errors recur across several products.
Finally, use the official sample questions to understand format, not as a source of undisclosed live exam material. Revisit the certification page before booking and the product references before using a command in production. The goal is to reason from service behavior, not memorize transient prices, obsolete model names or unqualified performance guarantees.
Sources: Certification and registration, current exam guide, retake policy, delivery transition. Checked October 18, 2026.
Use the free Professional Cloud Developer practice questions to test these decisions after studying the relevant chapter. Review the explanations and return to the teaching for the missed concept.
A candidate failed the professional exam twice. What current minimum wait and attempt limit apply?
14 days and four attempts per year.
60 days before the third attempt, with at most four attempts in two years.
365 days before the third attempt, with unlimited attempts.
No wait if the next attempt uses a different language.
A Cloud Run invoice handler returns success before saving the invoice, then loses work when an instance terminates. Which design meets durable completion requirements?
Increase logging verbosity only.
Assume minimum instances prohibit termination.
Queue durable work, use an idempotent worker and acknowledge after durable completion.
Use an in-memory flag to remember completed requests forever.
Sections you finish are checked off in the contents.