SnowPro Core Prep

Lab 19 · Capstone · You are Priya

Timed mock — 40 questions

COF-C03 style coverage across all five domains. No giant SQL — this is exam muscle. Full key also in MOCK_ANSWERS.md. Night-before tips: EXAM_DAY_PLAYBOOK.md.

Business problem Morgan booked Pearson VUE next month. Priya feels “mostly ready” but has never sat a full timed set. Guessing on Domain 5 and warehouse scale-out will waste a seat fee.
Solution you will build One honest timed mock: 45 minutes, 40 questions, score yourself /40. Need ≥32 (80%) before you book. Weak-spot map sends you back to the right labs.
Before
Hope-based booking
  • “I watched the videos”
  • No timed practice
  • Guess on edition traps
After
Evidence-based booking
  • ≥32/40 under time
  • Weak domains re-drilled
  • Playbook night before

Episode 19 · Capstone

Morgan: “Book after proof, not vibes.”

MorganManager
Forty questions. Timer on. No peeking at answers until you’re done. Eighty percent or we don’t book.
PriyaYou
Deal. I’ll mark wrong domains, re-say the drills aloud, then retake once cold.

Treat this like Pearson: one pass, then score. Peeking early trains false confidence.

Instructions

Time
45 minutes

Phone timer. Close other tabs. ~1 minute per question average.

Official exam is longer;
this is a focused mock
Scoring
≥32 / 40 (80%)

Write A/B/C/D on paper. Open each <details> only after the full set.

Also: MOCK_ANSWERS.md
for a clean key

Questions are grouped by domain. Each item lists four choices; expand Answer for the correct letter and why distractors fail.

Domain 1 — Architecture (Q1–Q12)

~31% of the real exam. Labs 01–05.

  1. Which statement about virtual warehouses is true?

    • A. Warehouses store micro-partitions for tables
    • B. Warehouses provide compute; table data lives in the storage layer
    • C. Larger warehouses always reduce storage cost
    • D. Cloud Services runs inside your warehouse size
    Answer

    B. WH = compute. Storage holds compressed micro-partitions. A confuses layers; C/D are wrong.

  2. Name the three Snowflake architecture layers.

    • A. Database, Schema, Table
    • B. Storage, Query Processing (warehouses), Cloud Services
    • C. Snowsight, SnowSQL, JDBC
    • D. Standard, Enterprise, Business Critical
    Answer

    B. Classic three-layer answer. A is object hierarchy; C interfaces; D editions.

  3. Which edition can configure Time Travel up to 90 days?

    • A. Standard only
    • B. Enterprise and above
    • C. Reader accounts only
    • D. Any edition with Fail-safe enabled
    Answer

    B. Standard defaults to 1 day; Enterprise+ for up to 90. Fail-safe ≠ TT length control.

  4. Micro-partitions are best described as:

    • A. Fixed 50 MB row groups you manually size
    • B. Contiguous compressed storage units Snowflake manages automatically
    • C. Warehouse cache blocks only
    • D. One partition per table always
    Answer

    B. Auto-managed compressed units (~16 MB uncompressed target historically). Not warehouse cache; not one-per-table.

  5. Which interface is the web UI for Snowflake?

    • A. SnowSQL
    • B. Snowsight
    • C. Kafka Connector only
    • D. Fail-safe console
    Answer

    B. Snowsight = web. SnowSQL = CLI.

  6. Which is NOT a table type category the exam expects you to distinguish?

    • A. Permanent
    • B. Transient
    • C. Temporary
    • D. Fail-safe table
    Answer

    D. Fail-safe is a retention period after TT — not a CREATE TABLE type.

  7. Cloud Services layer includes:

    • A. Only virtual warehouse CPU
    • B. Authentication, metadata, query optimization/management
    • C. Only S3/Blob customer buckets you create
    • D. Only Snowpipe billing
    Answer

    B. Cloud Services = control plane (auth, metadata, optimizer, etc.).

  8. Business Critical is most associated with which need?

    • A. Free reader accounts only
    • B. Higher security / some advanced failover & compliance features
    • C. Disabling Cloud Services
    • D. Removing Time Travel
    Answer

    B. BC for stricter security and many HA/failover capabilities.

  9. A secure view is used primarily to:

    • A. Force warehouse size to X-Small
    • B. Hide underlying details / prevent some optimization peeking at base objects
    • C. Replace Fail-safe
    • D. Enable Marketplace billing only
    Answer

    B. Secure views protect definition/underlying object exposure compared to normal views.

  10. Which cloud partners host Snowflake? (exam spirit)

    • A. Only on-prem VMware
    • B. Major public clouds (AWS, Azure, GCP)
    • C. Only AWS Gov alone forever
    • D. Customer laptop Docker only
    Answer

    B. Multi-cloud SaaS on AWS/Azure/GCP (regions/editions vary).

  11. Cortex / Snowpark / Streamlit in Snowflake sit closest to which blueprint theme?

    • A. Fail-safe administration
    • B. AI / apps / developer platforms (architecture awareness)
    • C. COPY INTO file formats only
    • D. Reader account billing only
    Answer

    B. Domain 1 AI/apps awareness on C03.

  12. Clustering keys help primarily with:

    • A. Making warehouses auto-suspend faster
    • B. Improving pruning / scan efficiency when filters match cluster dimensions
    • C. Granting ACCOUNTADMIN rights
    • D. Creating reader accounts
    Answer

    B. Clustering influences micro-partition organization for pruning — not RBAC or suspend.

Domain 2 — Account management & governance (Q13–Q20)

~20% of the real exam. Labs 06–09.

  1. RBAC in Snowflake is primarily about:

    • A. Assigning privileges to roles, then roles to users
    • B. Storing passwords in tables
    • C. Cloning warehouses
    • D. Fail-safe tickets
    Answer

    A. Classic role-based access: privileges → roles → users.

  2. Dynamic data masking typically:

    • A. Deletes columns permanently
    • B. Transforms column values at query time based on role/policy
    • C. Doubles warehouse size
    • D. Enables replication
    Answer

    B. Masking policies alter what roles see without rewriting stored data.

  3. Row access policies control:

    • A. Which rows a role can see
    • B. Warehouse auto-resume only
    • C. Fail-safe duration
    • D. Stage file format
    Answer

    A. RAP = row-level filter; masking = column transform.

  4. Resource Monitors are used to:

    • A. Mask emails
    • B. Track/limit credit usage and notify or suspend
    • C. Create shares
    • D. Replace Time Travel
    Answer

    B. Cost governance via credit quotas/actions.

  5. ACCOUNT_USAGE views are useful for:

    • A. Live millisecond operational locks only
    • B. Historical monitoring of queries, logins, usage (with latency)
    • C. Replacing INFORMATION_SCHEMA entirely for all DDL
    • D. Fail-safe restores
    Answer

    B. Account-level history; expect latency vs INFORMATION_SCHEMA.

  6. Network policies primarily restrict:

    • A. Which IP/CIDR ranges can authenticate
    • B. Micro-partition size
    • C. Marketplace prices
    • D. Clone privileges only
    Answer

    A. Network policies allow/block client IPs.

  7. Object tagging helps mainly with:

    • A. Governance classification / tracking sensitive data
    • B. Speeding COPY INTO always
    • C. Disabling Cloud Services
    • D. Creating Fail-safe
    Answer

    A. Tags for governance/classification workflows.

  8. MFA / SSO / key-pair auth are examples of:

    • A. Storage layer features
    • B. Authentication / access security controls
    • C. Clustering keys
    • D. Transient table options
    Answer

    B. Domain 2 security authentication themes.

Domain 3 — Loading, unloading & connectivity (Q21–Q27)

~18% of the real exam. Labs 10–12.

  1. COPY INTO a table from a stage is best described as:

    • A. Continuous auto-ingest only
    • B. Batch load you run (or schedule) from files
    • C. Fail-safe restore
    • D. Zero-copy clone
    Answer

    B. COPY = batch. Snowpipe = continuous/auto pattern.

  2. Snowpipe is designed for:

    • A. Continuous/near-continuous loading as files arrive
    • B. Replacing all warehouses
    • C. Manual UNDROP only
    • D. Marketplace billing only
    Answer

    A. Pipe automation vs batch COPY.

  3. A named stage can reference:

    • A. Only local laptop RAM
    • B. Internal Snowflake stages or external cloud storage locations
    • C. Fail-safe only
    • D. Reader accounts only
    Answer

    B. Internal/external stages hold file references for load/unload.

  4. Streams track:

    • A. Warehouse suspend events only
    • B. Change data (CDC-style) on a table for downstream processing
    • C. Marketplace listings
    • D. Fail-safe tickets
    Answer

    B. Table streams = change tracking for incremental pipelines.

  5. Tasks are used to:

    • A. Schedule SQL / pipeline steps in Snowflake
    • B. Mask columns
    • C. Create editions
    • D. Replace micro-partitions manually
    Answer

    A. Tasks schedule work (often with streams).

  6. File format objects define:

    • A. CSV/JSON/Parquet parsing options for load/unload
    • B. Role hierarchies
    • C. Resource monitor quotas only
    • D. Clustering keys only
    Answer

    A. FILE FORMAT = how files are interpreted.

  7. Storage integrations help primarily with:

    • A. Securely connecting Snowflake to cloud storage without embedding keys in every stage
    • B. Creating transient users
    • C. Disabling Time Travel
    • D. Cloning ACCOUNTADMIN
    Answer

    A. Integration object for external stage auth to cloud storage.

Domain 4 — Performance, querying & transformation (Q28–Q35)

~21% of the real exam. Labs 13–15.

  1. Result cache returns prior results when:

    • A. Always, even if underlying data changed
    • B. Query text/results are eligible and base data hasn’t changed (within rules)
    • C. Only on X-Small warehouses
    • D. Only in Fail-safe
    Answer

    B. Result cache is conditional — data change invalidates.

  2. Metadata cache / partition pruning mainly reduces:

    • A. Bytes scanned by skipping irrelevant micro-partitions
    • B. Role grants
    • C. Fail-safe days
    • D. Number of shares
    Answer

    A. Pruning = skip partitions that can’t match filters.

  3. Scale-up vs scale-out:

    • A. Scale-up = larger WH size; scale-out = more clusters for concurrency
    • B. Both mean adding reader accounts
    • C. Scale-out only exists on Standard
    • D. Scale-up deletes micro-partitions
    Answer

    A. Size vs multi-cluster concurrency — classic trap pair.

  4. Auto-suspend saves money by:

    • A. Deleting tables after idle
    • B. Stopping warehouse credit burn when idle past the threshold
    • C. Disabling Cloud Services
    • D. Clearing Fail-safe
    Answer

    B. Idle WH suspends → no compute credits while suspended.

  5. Query Profile is used to:

    • A. Diagnose operators, spills, pruning, bottlenecks
    • B. Create shares
    • C. Set MFA
    • D. UNDROP databases
    Answer

    A. Performance triage tool.

  6. VARIANT columns are for:

    • A. Semi-structured data (JSON-like) storage/query
    • B. Only integers
    • C. Warehouse sizes
    • D. Network policies
    Answer

    A. Semi-structured; often with FLATTEN / path syntax.

  7. QUALIFY is most associated with:

    • A. Filtering window-function results
    • B. Granting USAGE on shares
    • C. Enabling replication
    • D. Setting retention days
    Answer

    A. QUALIFY filters after window functions (like HAVING for aggregates).

  8. Bigger warehouse always faster?

    • A. Yes, always
    • B. No — if bottleneck is pruning/IO/spill/concurrency, Profile first; size may not help
    • C. Only if Fail-safe is on
    • D. Only for shares
    Answer

    B. High-frequency trap from the blueprint.

Domain 5 — Data collaboration (Q36–Q40)

~10% of the real exam. Labs 16–18.

  1. Time Travel vs Fail-safe:

    • A. Same thing; both are SELECT AT
    • B. TT is user-recoverable in retention; Fail-safe is Support-only after TT
    • C. Fail-safe is longer on transient tables
    • D. UNDROP only works in Fail-safe
    Answer

    B. Core Domain 5 trap. Transient has no Fail-safe; UNDROP is TT-window.

  2. Zero-copy clone storage charges begin when:

    • A. Immediately equal to a full second copy of all bytes
    • B. Objects diverge via writes creating new micro-partitions
    • C. You run SHOW SHARES
    • D. Fail-safe starts
    Answer

    B. Shared pointers until divergence.

  3. Secure Data Sharing is best described as:

    • A. Nightly CSV dump to the partner
    • B. Live access to provider data via share grants (not a physical copy)
    • C. Cloning into Fail-safe
    • D. Disabling RBAC
    Answer

    B. Live share, not export.

  4. Who typically pays compute when a consumer queries a share?

    • A. Always Snowflake Support
    • B. The consumer’s warehouse credits
    • C. Fail-safe pool only
    • D. Only ACCOUNTADMIN on provider
    Answer

    B. Consumer compute; provider storage of shared data.

  5. A reader account is:

    • A. A provider-managed account for consuming shared data without a full customer account
    • B. A type of micro-partition
    • C. Synonym for Fail-safe
    • D. Only available on temporary tables
    Answer

    A. Reader = limited consumer account created by provider.

Score yourself

Pass bar
≥32 / 40

Below 32: do not book. Re-drill weak domains, then retake cold (different day).

Official pass is scaled 750;
use 80% here as a gate
Key
Quick letter list

Also open MOCK_ANSWERS.md for the full key + distractor notes.

1–12 Domain1 · 13–20 D2
21–27 D3 · 28–35 D4 · 36–40 D5

Weak-spot map — wrong domain → go here

D1 misses
Architecture

Revisit Labs 01–05. Re-say three layers, editions, micro-partitions aloud.

Q1–Q12 wrong → Lab 01 first
D2 misses
Security & cost

Labs 06–09: RBAC, masking vs RAP, monitors, ACCOUNT_USAGE.

Q13–Q20 → Lab 06 + 07
D3 misses
Load / pipelines

Labs 10–12: COPY vs Snowpipe, stages, streams/tasks, integrations.

Q21–Q27 → Lab 10 + 11
D4 / D5
Perf + collaboration

Labs 13–15 performance; Labs 16–18 TT/clone/share. Blueprint trap table.

Q28–Q35 · Q36–Q40

Exam-day strategy drills — say these out loud

Q1. First 5 minutes in the Pearson seat — what do you do?

Say: Breathe, confirm NDA/rules, note rough time budget (~1+ min/question on the real exam), flag hard items mentally, don’t spiral on Q1. Process over panic.

Q2. Two answers look right — how do you choose?

Say: Eliminate using architecture layers, edition limits, and “who pays / who recovers.” Prefer the option that matches Snowflake’s separation of storage/compute/services. Mark and move if still stuck.

Q3. You blank on Fail-safe — recovery line?

Say: Time Travel = me + SQL inside retention. Fail-safe = after TT, Support only, ~7 days permanent. Transient ≠ Fail-safe. That triad unlocks most Domain 5 recoveries.

Q4. Performance question mentions “slow” — first mental checklist?

Say: Pruning/filter selectivity → bytes scanned → spill → warehouse size vs multi-cluster concurrency → caches. Don’t jump to “always scale up.”

Q5. When do you book the real exam?

Say: After ≥32/40 on this mock (or equivalent), weak domains re-drilled aloud, and playbook skimmed night before — not after a single lucky peek at answers.

Priya’s wrap-up — booking checklist
“I timed 40 questions without peeking. Score ___ / 40. If ≥32: skim EXAM_DAY_PLAYBOOK.md, re-say Domain 5 traps, book Pearson VUE. If <32: weak-spot map → labs → cold retake. Checklist: ID ready · quiet room / test center plan · MFA know-how for Snowflake jobs not for the exam login · sleep · no new topics morning-of. Outcome: book from evidence, not vibes.”
Also open Answer key · Exam day playbook · Blueprint checklist
← Lab 18 Home →