2.5 Practice Drills and Readiness Markers
Key Takeaways
- Build a layer map (capacity / domain / workspace / item) and place every control and symptom on it from memory.
- Drill the ALM split: Git = inner loop / definitions / one branch; deployment pipelines = outer loop / stages / no data, no credentials.
- Memorize the role matrix (Admin > Member > Contributor > Viewer) by the verb each unlocks: manage, share, edit, view.
- Event-based triggers (file arrival) and schedule triggers are different; pick by whether timing is event-driven or clock-driven.
- You are ready when you can name the layer, the control, and why the tempting distractor is wrong for mixed questions after a day off.
Drill 1: The Layer Map
Draw four boxes — capacity, domain, workspace, item — and, without notes, write every control and every common symptom into the right box. A correct map looks like this:
| Layer | Controls | Tell-tale symptom |
|---|---|---|
| Capacity (F SKU) | CUs, throttling, autoscale, pause | "Free viewers can't see content" (need F64) |
| Domain | Federated governance, domain admins, catalog discovery | "Business unit should self-govern" |
| Workspace | Roles, Git binding, deployment stage | "Who can edit/share/manage" |
| Item | Item permissions, OneLake security, sharing, labels | "Read one lakehouse / file access" |
If you place a symptom in the wrong box during the drill, that is exactly the miss the exam will exploit. Re-derive the map daily until it is automatic.
Drill 2: ALM and the Role Matrix
Recite the ALM split aloud: Git integration is the inner loop — one workspace bound to one branch in Azure DevOps or GitHub, storing definitions (warehouse = SQL database project), never data. Deployment pipelines are the outer loop — up to 10 stages, promoting item definitions Dev->Test->Prod, using deployment rules and parameters to swap stage-specific values, and never copying data source credentials or table data.
Then drill the role matrix by the verb it unlocks:
- Viewer -> view only
- Contributor -> view + edit
- Member -> view + edit + share + create OneLake security roles
- Admin -> everything, including permissions, Git, and deployment settings
For every practice miss, write "I missed this because..." naming the wrong layer, the wrong ALM mechanism, or the over-granted role. A clear cause turns a miss into a recognizable cue.
Drill 3: Orchestration and Triggers
Readiness here means instantly choosing the right orchestration object. Use a two-column sheet: cue on the left, control on the right.
| Cue in the stem | Correct choice |
|---|---|
| Coordinate steps, schedule, retry | Data pipeline |
| Code-heavy Spark transformation, custom libraries | Notebook (invoked by the pipeline) |
| Start work when a file arrives | Event-based (storage event) trigger |
| Start work at a fixed time | Schedule trigger |
| Reproducible Spark runtime/libraries across stages | Environment item |
A domain is ready when, after a one-day break, you can answer mixed questions without the domain label, name the layer and control, and explain why the convenient distractor (a higher role, a copy that "moves" data, the wrong endorsement tier) is wrong. If your score drops sharply after the break, your knowledge is recognition-based — switch to active recall on the layer map and the ALM split until it holds.
Drill 4: Governance and Security Rapid-Fire
Governance facts are dense, so drill them as one-line cue-to-control pairs until they are reflexive:
- Federated, self-governing business units -> domains + delegated domain admins.
- Trusted but self-service endorsement -> Promoted; official, reviewer-vetted -> Certified; golden single source -> Master data.
- Classify and protect items, inherits to children -> Purview sensitivity labels.
- Native governance insights, no Purview required -> OneLake catalog Govern tab.
- Cross-platform discovery, DLP, enterprise classification -> Microsoft Purview.
- Row/column/folder data access -> OneLake security roles (Admin or Member creates them).
- Read one item without workspace access -> share the item.
Drill 5: Build and Grade Your Own Stems
The highest-value drill is writing your own scenario from a single fact, then grading it. Take "data source credentials are not copied in deployment," and author a Test-to-Prod promotion question whose distractors are all definitions are copied, reports are copied, and measures are copied. Forcing yourself to invent plausible-but-wrong options trains you to spot them on the real exam. Do the same for the F64 threshold, the one-branch Git rule, the pipeline-versus-notebook split, and the Viewer-needs-a-data-access-role pattern.
Write a one-sentence explanation for any concept you could not cleanly justify, naming the layer, the mechanism, or the over-granted role. When you can generate and defend a question for every high-yield fact in this domain, you have moved from recognition to genuine, transferable mastery — the readiness marker the DP-700 administration domain is testing.
Drill 6: The Numbers and Names You Must Know Cold
DP-700 administration leans on a small set of exact facts; convert them into a flash sheet and recite it until perfect. F64 is the capacity threshold for free-viewer consumption and Power BI Premium parity. Capacity Units (CUs) are the compute unit, and SKUs name their CU/second allowance (F2, F8, F64, up to F2048). Deployment pipelines support up to 10 stages. Git integration supports exactly two providers (Azure DevOps Git, GitHub) and one branch per workspace. There are four workspace roles (Admin, Member, Contributor, Viewer). Endorsement has three tiers (Promoted, Certified, Master data).
Only Admin or Member can create OneLake security roles. A warehouse serializes to a SQL database project in Git. If you can state each of these without hesitation, name the layer it belongs to, and produce one scenario where it is the deciding fact, you are ready for the administration domain. Schedule a final mixed review the day before the exam covering only these numbers and names, because they are the cues that disambiguate the hardest two-answer questions.
You want a Fabric process to start automatically when a new file lands in storage, rather than at a fixed time. Which trigger type fits?
Which workspace role is required to manage the workspace's Git connection and deployment pipeline settings?
A transformation needs custom Python libraries and distributed Spark execution across a very large dataset. Which item should perform the core transformation?