Cloudera Data Lakehouse Engineer Exam: Where the 60 Questions Go

The Cloudera Data Lakehouse Engineer exam is a 60-question online test that you finish in 90 minutes and pass with a score of 60%. Its questions come from three areas: Apache Iceberg table architecture (40%), Apache Ozone storage (35%), and distributed data operations such as distcp and the Ozone shell (25%).
The exam is new to Cloudera's certification list, so there is little written about it yet beyond Cloudera's own page. This note takes that page and turns the weights into numbers you can study with: how many questions each topic carries, how many you can miss, and where your study weeks should go. All facts were checked on Cloudera's pages on 29 October 2026.
What does the Data Lakehouse Engineer exam test?
It tests whether you can design, run and repair an open lakehouse on the Cloudera platform. In practice that means two technologies and the tools around them. Iceberg is the table format that gives files on storage the behaviour of database tables: snapshots, safe concurrent writes, schema and partition changes. Ozone is the object store those files can live on. The third domain covers moving data at scale and asking the storage layer about its own state.
The official Data Lakehouse Engineer exam page lists each domain with its weight and a detailed list of objectives. That list is the real syllabus, and the rest of this note follows it closely.
What are the exam facts?
Cloudera gives the format in a short block at the top of the exam page. The retake and validity rules sit in the program FAQ. Here they are side by side.
| Item | What Cloudera states |
|---|---|
| Questions | 60 |
| Time | 90 minutes |
| Pass score | 60% |
| Delivery | Online, proctored (Questionmark); no test centre option |
| Allowed resources | None: no notes, white papers or user guides |
| Results | On screen and by email right after you submit, with topic results |
| Retake | Wait 7 calendar days after a fail; every attempt is paid in full |
| Validity | 2 years; you may sit the exam again once it expires |
| Price | Not shown publicly; the store asks for a Cloudera sign-in first |
Two points are easy to miss. The exam page gives no exam code, so search results that attach an older code to this exam are guessing. And the price is only visible inside Cloudera's education store after you sign in, so check it there before you plan a budget.
How many questions does each domain carry?
Cloudera publishes percentages, not question counts. With 60 questions the maths is clean, because each weight turns into a whole number.
| Domain | Weight | Questions (of 60) | Time at 90 s each |
|---|---|---|---|
| 1. Transactional open table architecture (Apache Iceberg) | 40% | about 24 | 36 min |
| 2. High-scale enterprise storage layouts (Apache Ozone) | 35% | about 21 | 31.5 min |
| 3. Distributed administrative utilities and data operations | 25% | about 15 | 22.5 min |
| Total | 100% | 60 | 90 min |
The real exam may not match these counts exactly, but they show where the weight sits. Three things follow from them:
- You need 36 correct answers. That leaves room for 24 misses, which is generous, but only if the misses are spread out.
- Iceberg alone cannot pass you. A perfect Iceberg score is about 24, still 12 short of 36. Ozone and the utilities together carry 36 questions, which is 60% of the exam.
- 90 seconds a question is enough time to read a command. Domains 2 and 3 ask you to spot a missing part of a path or pick the right shell syntax. Those are quick if you have typed the commands yourself, and slow guesses if you have only read about them.

Does the short summary match the scored blueprint?
Not quite, and this matters for how you study. Cloudera's certification list describes the exam in a paragraph, while the exam page gives the scored objectives. The two stress different things.
| Topic | In the summary paragraph | In the scored blueprint |
|---|---|---|
| Iceberg snapshots, rollback, partition evolution, hidden partitioning | Yes | Yes, in detail (40%) |
| Apache Ozone (Ozone Manager, volumes, buckets, keys, ofs://) | Not named | Yes, a full domain (35%) |
| hadoop distcp and the ozone sh shell | Not named | Yes (25%) |
| Spark, Trino and Hive query tuning, compaction | Yes | Not listed as an objective |
The Cloudera certification program list is where the summary paragraph lives. If you studied from that paragraph alone, you might spend weeks on engine tuning and never open an Ozone shell. Study from the exam page objectives, and treat the engine material as background.
Which Iceberg ideas carry the 40%?
The Iceberg domain is about how the table format works inside, not about writing SQL. The objectives group into five ideas.
The metadata tree. A table's metadata JSON file points to its history. Each snapshot has a manifest list, and each manifest list points to manifest files. Those manifest files record data file paths and column-level metrics. Expect questions that ask where a given statistic lives, for example row counts, null counts or counts of not-a-number values. The exam contrasts statistics kept in manifests with those read from raw file footers.
Why planning is fast. Because file-level summaries sit in the metadata, a query planner can rule out files without listing folders. The exam frames this as constant-time planning against the older linear-time directory scan. You should be able to explain the difference in one sentence.
Deletes. Copy-on-Write rewrites the affected data files at delete time. Merge-on-Read writes a small delete record and applies it when the table is read. Know which one makes writes cheaper and which makes reads cheaper, and when each fits.
Partitioning. With hidden partitioning, Iceberg works out the partition from a column value, so users filter on the column itself and still get pruning. Partition evolution lets the layout change over time without rewriting old data.
Safety patterns. Write-Audit-Publish stages new data, checks it, then makes it visible. Time travel reads an older snapshot for an audit. A snapshot rollback restores the whole table at once.
The Apache Iceberg table specification describes the metadata files, manifests and delete files in full. Read its overview and the sections on manifests and row-level deletes, not every appendix.
What should you know about Ozone and the operations tools?
Ozone carries 35% and is the domain most candidates from a pure Spark background will know least. The objectives focus on three things.
- The Ozone Manager. It handles the namespace and keeps it in RocksDB on disk, holding only the active part of the metadata in memory. That is how it scales to very large numbers of objects.
- The hierarchy. Volumes are administrative boundaries, buckets hold a user's storage, and keys are the individual objects. Expect to tell these three boundaries apart.
- Paths. Through the ofs:// interface, a full path names the volume and then the bucket before the key. One objective asks you to spot the part that is missing from such a path.
Cloudera's Apache Ozone overview in the Runtime documentation covers the architecture, containers and how reads, writes and deletes are handled. It is the closest official match to this domain.
The last 25% is about data operations. You need to know why a single-threaded copy is slow between storage systems, and how distcp splits the job across the cluster to run in parallel. You also need the ozone sh command families, such as volume info, to query a volume or bucket. The Apache Hadoop DistCp guide explains its options and how the copy is divided into map tasks.
How should you plan four weeks of study?
Split your time the way the exam splits its questions, and make every week end with hands-on work. This plan assumes about 8 to 10 hours a week and some prior Hadoop or Spark experience.
| Week | Focus | Hands-on check |
|---|---|---|
| 1 | Iceberg metadata tree, planning, statistics | Create a table, insert twice, and find the two snapshots in its metadata |
| 2 | Deletes, hidden partitioning, partition evolution, WAP, time travel, rollback | Delete rows, change the partition spec, then roll back to the first snapshot |
| 3 | Ozone Manager, volumes, buckets, keys, ofs:// paths | Create a volume and bucket, write a key, and read it back through a full ofs:// path |
| 4 | distcp, ozone sh, then a full review against the objective list | Copy a folder with distcp, then query the target with ozone sh volume info |
Weeks 1 and 2 cover Iceberg because it is the largest single domain. Weeks 3 and 4 carry the other 60% of the questions, so do not let Iceberg spill into them. If you have no cluster, Cloudera suggests two optional courses that match the domains: DENG-255 for Iceberg and ADMIN-237 for Ozone.
For the Iceberg side as Cloudera runs it, the Iceberg overview in Cloudera Runtime shows which engines and catalog the platform uses by default.
What happens on exam day?
You book and sit the exam online through Questionmark, and a proctor watches by webcam and screen share. The Cloudera certification FAQ sets out the rules:
- Run the system test and install Zoom about 48 hours before, and use a wired connection if you can.
- You must be alone, silent and fully visible. No notes, phones, headphones, food, breaks or other applications.
- If your connection drops, you can rejoin and continue, but the timer keeps running.
- You can cancel or move a booking up to 24 hours before the start. A no-show loses the fee.
Your result appears on screen as soon as you submit, with topic-level results. Use those topic results to plan a retake if you need one. You cannot sit the exam again for 7 days, and the next attempt costs the full fee again.
Is there a video on preparing for a Cloudera data exam?
Our channel has no video on the Lakehouse exam yet. This short one covers how to prepare for Cloudera's Data Engineer exam, the closest exam in the same track. Its advice on working from the official objective list fits this exam too.
Frequently asked questions
1. Is the Data Lakehouse Engineer exam the same as the Cloudera Data Engineer exam?
No. They are listed as separate certifications. The Data Engineer exam covers the data engineer role on the Cloudera platform, while the Lakehouse exam is about the table format and storage layer underneath: Iceberg, Ozone, distcp and the Ozone shell.
2. Can I use the Iceberg or Ozone documentation during the exam?
No. Cloudera lists the allowed resources as none. Reference material, white papers and user guides are all excluded, and you may not open any website other than the exam itself.
3. How long is the certification valid?
Two years. When it expires, you may take the exam again to keep the certification. You cannot retake it while it is still valid just to raise your score.
4. Do I have to take Cloudera training before I book?
No. Cloudera calls its suggested courses optional. It does recommend real hands-on experience, which matters most for the Ozone path and shell questions.
5. Can I take the exam at a test centre?
Not at present. Cloudera says its certification exams are available online only, through Questionmark with a live proctor.
If you are weighing Cloudera’s data and AI tracks, our companion note on the Cloudera Generative AI Engineer exam breaks that exam’s 60 questions down the same way, from LLMs and RAG to the Cloudera platform work that fills nearly half of it.
Last updated 29 October 2026. Exam facts checked on Cloudera's exam page, certification list and certification FAQ that day.
Written by the team behind AnalyticsExam.
Comments
Post a Comment