GCP Cloud DevOps Engineer (PCDOE): what's actually on the exam
PCDOE tests SRE and DevOps practices on GCP — Cloud Build, Cloud Deploy, GKE pipelines, SLO design, incident response. Here's what to expect and how it compares to AZ-400 and DOP-C02.
PCDOE — Professional Cloud DevOps Engineer — is Google's SRE-flavored DevOps cert. $200, two hours, around 50 questions, validity two years. If you've taken any Google Professional exam, that's the standard shape. What's not standard is the case-study density. PCDOE has the reputation of being one of the harder Google pro exams to pass on the first attempt, and the reason is almost entirely that the questions are long, scenario-heavy, and assume you've actually run a service in production.
Pass rates aren't published — Google doesn't publish numeric scores either, just pass/fail — but the anecdotal first-attempt rate from study group threads sits noticeably below ACE and a touch below PCA. Take that with the usual salt. Self-reported pass rates skew toward people who failed and want to vent.
What the exam actually covers
The official guide breaks PCDOE into five domains. The weight gets reshuffled every couple of years, but the shape is consistent:
- Applying site reliability engineering principles. SLO and SLI design, error budgets, toil reduction, blameless postmortems. Direct lift from Google's SRE book — read at least the chapters on SLOs, alerting, and incident response if you haven't.
- Building and implementing CI/CD pipelines. Cloud Build, Artifact Registry, Cloud Deploy, Skaffold for local-to-cluster workflows, integration with GitHub / GitLab. Container scanning with Artifact Analysis. Binary Authorization for signed images.
- Implementing service monitoring strategies. Cloud Monitoring, Cloud Logging, Cloud Trace, Cloud Profiler, Error Reporting. The whole "Cloud Operations" suite (formerly Stackdriver — the docs still say "Cloud Ops" but recruiters and old engineers will use both names interchangeably).
- Optimizing service performance. GKE workload tuning, autoscaling (HPA, VPA, cluster autoscaler), cost optimization, capacity planning.
- Managing service incidents. On-call rotations, incident command, runbooks, postmortem writeups. Yes, they ask process questions, not just tool questions.
The case-study questions are the hard part. You'll get a scenario describing a fictional company — its architecture, its current pain points, its team structure — and three or four questions that require you to hold all of that in your head simultaneously. Skim-reading is how people fail. The questions are designed so that the obvious answer becomes wrong once you re-read the scenario carefully.
What you actually need to know
Not every GCP service shows up. Here's the rough hit list, weighted by how often topics appear in study reports:
| Service / topic | Weight on exam |
|---|---|
| Cloud Build, Cloud Deploy, Artifact Registry | Heavy |
| GKE operations, autoscaling, workload identity | Heavy |
| SLO / SLI / error budget math | Heavy |
| Cloud Monitoring, alerting policies, dashboards | Heavy |
| Cloud Logging, log-based metrics, log routing | Medium |
| Binary Authorization, container scanning | Medium |
| Cloud Trace, Profiler, Error Reporting | Medium |
| Terraform / Config Connector basics | Medium |
| Anthos / multi-cluster (lighter than it used to be) | Light |
| Pub/Sub, Cloud Tasks for async patterns | Light |
You don't need to memorize every Cloud Build YAML key. You need to recognize when Cloud Build is the right answer vs Cloud Deploy vs a third-party CI plus Cloud Deploy. The exam loves "company X is using Y, what should they do next" framing.
How it compares to AZ-400 and DOP-C02
All three certs cover similar ground at the conceptual level — pipelines, monitoring, incident response, IaC, security — but with different emphases.
| PCDOE | AZ-400 | DOP-C02 | |
|---|---|---|---|
| Cost | $200 | $165 | $300 |
| Length | ~2h, ~50 q | ~150 min, ~50 q | ~3h, ~75 q |
| Validity | 2 years | 1 year, free renewal | 3 years |
| SRE / SLO depth | Heavy | Light | Medium |
| Native CI/CD focus | Cloud Build / Deploy | Azure DevOps + GitHub | CodePipeline / CodeBuild |
| IaC focus | Terraform, Config Connector | Bicep, ARM, Terraform | CloudFormation, CDK |
| Hardest part | Case-study density | Breadth of Azure DevOps | DOP-C02 wordy questions |
PCDOE leans the most on SRE concepts — SLOs, error budgets, toil — because Google literally invented that vocabulary. AZ-400 leans on Azure DevOps (the product) and GitHub Actions integration. DOP-C02 covers the widest surface area but in shallower scenario depth than PCDOE.
If you work in GCP, PCDOE is the obvious pick. If you work in Azure, AZ-400. If you work in AWS, DOP-C02. The skills overlap heavily and the pipelines look almost the same once you squint past the service names. Cert-shopping across clouds for DevOps is rarely the right move — the cert that maps to your day job will be the one you can actually study for using your real work.
Who it's for
Honestly, three cohorts:
Senior SREs / DevOps engineers already on GCP. This is the audience the exam is written for. If you've been on call for a service running on GKE for at least a year, the questions will feel like extensions of arguments you've already had in your team. Three to six weeks of focused prep is usually enough.
Platform engineers crossing from AWS or Azure. The SRE concepts transfer one-to-one. The service names don't. Expect 2-3 months of study to map your existing knowledge onto Cloud Build, Cloud Deploy, and the Cloud Ops suite. Build a small project that does CI through Cloud Build, deploys to GKE via Cloud Deploy, and emits SLOs to Cloud Monitoring. That single project covers maybe 40% of the exam.
Career switchers chasing a DevOps title. Honestly, this is a stretch goal. PCDOE assumes you've been through real production incidents. If you haven't, the case-study questions will look unfamiliar in a way that no amount of video courses will fix. Do CKA or KCNA first, work in a platform role for a year, then come back to PCDOE.
Study notes from the field
A few things that consistently trip people up:
The SRE book is required reading, not optional. Google's Site Reliability Engineering and The Site Reliability Workbook are both free online. The SLO / error budget chapters are directly testable. Skipping them and relying on courseware is the most common reason smart engineers fail this one.
Cloud Deploy is newer and gets disproportionate weight. It went GA in 2022. Earlier study guides under-cover it. Spend a weekend doing the official Cloud Deploy quickstart end-to-end with two environments and a canary rollout — that maps directly to multiple exam questions.
Know the difference between Cloud Build private pools and default pools. When questions involve VPC-internal builds, private pools are usually the answer. Generic "we want faster builds" questions are usually about machine type or worker pool size.
Binary Authorization shows up more than you'd expect. The "we need to enforce signed images in prod" scenario is almost guaranteed to appear in some form.
Bottom line
PCDOE is a solid pro cert if you actually do DevOps on GCP. It pays roughly the same band as PCA in similar markets — $150k–$200k base for senior DevOps / SRE roles in major US metros, with FAANG and ad-tech total comp pushing $250k+ once equity stacks. The cert isn't a salary multiplier on its own; the underlying SRE experience is. PCDOE just makes the experience legible to recruiters.
If you're studying, browse the PCDOE question bank on CertLabPro or start a timed exam. The case-study questions in the bank are the closest match to the real exam shape — straight-recall question dumps don't prep you for what actually shows up.
If you're deciding whether to bother: do you write postmortems at work? If yes, this cert will feel obvious. If no, get through one or two real incidents first, then come back.