> Doodling
← All 50 Days
Day 7 of 50
Review Week 1
DAY 07

Week 1 Recap, Drill + Exam Gotchas

No new material today: six days in, you have most of Domain 1’s foundation (architecture, four editions, interfaces, 12 database objects, parameter precedence). Today is a 30-minute closed-book drill, the recurring exam traps, and the Week 1 Practice Test, targeting 8 out of 10 to clear Week 2.

🗣️ How to use today
StepTimeWhat you do
1. Closed-book drill30 minRun the four drill tasks below. No tabs open. Write SQL on paper or a scratch buffer.
2. Self-grade5 minOpen Snowsight, run your SQL, score yourself honestly.
3. Read the gotchas10 minSpeed-read the patterns below. These are the recurring traps for Domain 1.
4. Take the Week 1 Practice Test15 min10 mixed-domain questions. Target: 8/10. Below that, re-read the day flagged in your wrong answers before moving on.
🛠️

The 30-Minute Drill

Type: CLOSED-BOOK DRILL  |  Time: ~30 minutes  |  Credits: <0.05  |  Rule: No docs, no Snowsight, no Day 1–6 tabs. Write the answers first, run them after.
1

Draw the 3-layer architecture from memory. On paper or a whiteboard, label each layer and write one example of work that runs in each.

CHECK
Expected:
  ┌─────────────────────────┐
  │   Cloud Services        │ ← auth, RBAC, optimizer, metadata
  │   (SHOW, DESCRIBE)      │   (no warehouse needed)
  ├─────────────────────────┤
  │   Compute (Warehouses)  │ ← MPP execution
  │   (SELECT, DML, MERGE)  │   (independent clusters)
  ├─────────────────────────┤
  │   Storage               │ ← columnar micro-partitions
  │   (immutable, shared)   │   (one copy, all warehouses)
  └─────────────────────────┘
👀 Self-check: Writing “data” for the Storage layer is not enough. The exam expects you to know it is immutable micro-partitions, columnar, compressed. If those three words did not come out, re-read Day 2.
2

Write context SQL from memory. One statement that returns your account name, current role, current warehouse, current region, current Snowflake version, current database, and current schema.

SQL
-- Try without peeking, then run this to verify:
SELECT
  CURRENT_ACCOUNT()                 AS account,
  CURRENT_ROLE()                    AS role,
  CURRENT_WAREHOUSE()               AS warehouse,
  CURRENT_REGION()                  AS region,
  CURRENT_VERSION()                 AS version,
  CURRENT_DATABASE()                AS db,
  CURRENT_SCHEMA()                  AS schema;
👀 Self-check: Forgetting CURRENT_REGION() or CURRENT_VERSION() on the first attempt is normal. They come up rarely in day-to-day work. Forgetting CURRENT_ROLE() is a red flag. Role context drives every governance question on the exam; re-skim Day 6.
3

List 5 features that are in Enterprise but NOT in Standard. No internet, no docs, no scrolling.

CHECK
Acceptable answers (any 5 of these):
  • Multi-cluster virtual warehouses
  • Time Travel up to 90 days (Standard caps at 1 day)
  • Dynamic Data Masking
  • Row Access Policies
  • Materialized Views
  • Search Optimization Service
  • Automatic Clustering (auto-clustering)
  • Object Tagging
  • Column-level Security
  • ACCESS_HISTORY view (governance)

NOT acceptable (these are in ALL editions):
  • Encryption (AES-256 at rest, TLS in transit)
  • Network policies
  • MFA, SSO, Federated Auth
  • Fail-safe (7 days)
  • Resource monitors
  • Object-level access control (RBAC)
👀 Self-check: Tri-Secret Secure and PrivateLink belong to Business Critical, not Enterprise. This is the single most common edition mistake I see in mock exams. The candidate knows the feature exists, places it one tier too low, and loses an easy mark. Re-read Day 3.
4

State the parameter precedence rule in one sentence, then resolve a 3-layer scenario.

CHECK
Rule (one sentence):
  Most-specific scope wins.
    Session params : Account → User → Session
    Object params  : Account → Database → Schema → Object

Scenario:
  STATEMENT_TIMEOUT_IN_SECONDS:
    Account = 3600
    User    = 1800
    Session = 600
  Effective value? _______

  Now: ALTER SESSION UNSET STATEMENT_TIMEOUT_IN_SECONDS;
  Effective value? _______ (in this session)

Answers:
  1) 600  — Session is most specific.
  2) Falls back to ACCOUNT (3600) in the current session.
     User-level overrides only kick in for NEW sessions.
👀 Self-check: Most candidates answer 1800 for the second part. The User-level value is set, so the intuition says “User is more specific than Account, pick User.” The catch is that User-level session parameters apply at login. The current session was already open when you ran UNSET, so it falls back to the account default. The exam tests this exact nuance.
5

Bonus speed round (60 seconds each):

CHECK
Q: Name 6 of the 12 database objects.
A: Any 6 of: Stages, Schemas, Tables, Views, UDFs, File Formats,
   Stored Procedures, Pipes, Shares, Sequences, ML Models, Applications.
   (Schemas are themselves a database object — yes, that's odd.)

Q: Which schema gives data with 0 latency but no dropped objects?
A: INFORMATION_SCHEMA. (ACCOUNT_USAGE has 45min–3hr latency
   and DOES include dropped objects.)

Q: Which two interfaces CAN'T run PUT or GET?
A: Snowsight worksheet UI cannot. PUT/GET need a CLI/driver
   (SnowSQL, Snowflake CLI, JDBC/ODBC drivers).

Q: Can you change the cloud (AWS↔Azure) for a Snowflake account?
A: No. Cloud and region are permanent per account.
   To “move” you create a new account in the target cloud
   and use replication.
👀 Self-check: 3 out of 4 right means you are tracking. 1 out of 4 means stop here. Spend 30 minutes back in the days you slipped on, then take the practice test. Taking the test cold after a weak drill wastes the test.
🎯

Week 1 Exam Gotchas

These are the patterns the exam reuses in scenario form. Each one has a Tell, the trigger phrase the question uses to set the trap. Learn the Tells and you will catch the trap before you read the options.

Gotcha 1: “Snowflake is a cloud provider” → FALSE

Snowflake is SaaS that runs on AWS, Azure, and GCP. It uses their infrastructure; it is not itself a cloud provider. Tell: any option that lists Snowflake alongside AWS, Azure, and GCP as a peer.

Gotcha 2: “Need a warehouse to run SHOW DATABASES” → FALSE

Metadata operations (SHOW, DESCRIBE, and some COUNT(*)/MIN/MAX queries that hit the metadata cache) are answered by the Cloud Services layer. No warehouse is needed. Tell: “all warehouses suspended, the query still returned instantly.”

Gotcha 3: Encryption is NOT edition-specific

AES-256 at rest and TLS in transit are always-on in every edition, Standard included. What is edition-specific is customer-managed keys via Tri-Secret Secure, which is Business Critical and higher. Tell: “What is the minimum edition needed for encryption?” The answer is Standard. The question is designed to make you reach for Business Critical.

Gotcha 4: PrivateLink is Business Critical, not Enterprise

Enterprise features are mostly analytical: multi-cluster, masking, materialized views, 90-day Time Travel. Connectivity and compliance features (PrivateLink, HIPAA/HITRUST, Tri-Secret Secure, account failover/failback) live one tier higher in Business Critical. Tell: any healthcare or PHI scenario, any “private network” or “regulated industry” framing.

Gotcha 5: VPS isolates hardware, not just network

VPS provisions a separate Snowflake environment with no shared hardware resources outside your account. Business Critical does not go that far. Tell: “completely isolated from all other Snowflake customers.” That phrasing only fits VPS.

Gotcha 6: PUT and GET require a client

PUT and GET move files between your local machine and a Snowflake stage. They need local file system access, which a browser cannot provide. The Snowsight worksheet UI cannot run them. Use Snowflake CLI, SnowSQL, or a driver. Tell: multi-select questions asking “which of these can run from a Snowsight worksheet.”

Gotcha 7: SnowSQL is not deprecated; it coexists with Snowflake CLI

Snowflake CLI (snow) is the newer tool. It covers Streamlit, Snowpark, Native Apps, and container services, where all new feature investment is going. SnowSQL is still supported for SQL and data loading workflows. Tell: “is SnowSQL deprecated?” The answer is no. The trap is the word deprecated.

Gotcha 8: Cloud + region is permanent per account

There is no “migrate account from AWS to Azure” setting. The supported path is: create a new account in the target cloud and region, then use database or account replication. Tell: “the team wants to switch the account from us-east-1 to eu-west-1.”

Gotcha 9: INFORMATION_SCHEMA does not show dropped objects; ACCOUNT_USAGE does

If the scenario asks “how many tables existed last quarter, including ones that have been dropped,” the only correct source is ACCOUNT_USAGE. INFORMATION_SCHEMA is real-time but only shows currently-existing objects. Tell: the words dropped, historical, or last year in the prompt.

Gotcha 10: ACCOUNT_USAGE latency varies from 45 min to 3 hr, not a single number

Most views land around 2 hours. Some run longer (METERING_HISTORY can be up to 3 hours, with a few columns even slower). Tell: any answer option that pins the latency to a flat number. It is almost always the wrong one.

Gotcha 11: Most-specific parameter scope wins

Account-level never overrides Session or Object level. The hierarchy walks from broad to narrow. The narrowest level that has a value wins. UNSET at one level reveals the next-broadest level that is still set above. It does not reset to the default. Tell: three-layer scenarios with mixed values, usually built around STATEMENT_TIMEOUT_IN_SECONDS or DATA_RETENTION_TIME_IN_DAYS.

Gotcha 12: DATA_RETENTION uses the Database/Schema/Object lane, not the Session lane

Object parameters cascade Account → Database → Schema → Object. Session parameters cascade Account → User → Session. The two lanes do not mix. Tell: any DATA_RETENTION_TIME_IN_DAYS scenario that drops a User-level value into the setup. The User-level value is a distractor. DATA_RETENTION does not flow through User.

Gotcha 13: The “12 database objects” multi-select trap

The exam lists 6–8 items and asks “which are database objects?” The wrong-column traps are always the same: Roles, Users, Warehouses, Resource Monitors, Network Policies, Storage Integrations. Those are account-level securables, not database objects. The 12 database objects are Stages, Schemas, Tables, Views, UDFs, File Formats, Stored Procedures, Pipes, Shares, Sequences, ML Models, and Applications.

Week 1 Readiness Checklist

If you can confidently…You’re solid onOtherwise re-skim
Name the 3 architecture layers and what runs in eachDay 1, Day 2Day 2: 3-layer architecture
Match HIPAA, PrivateLink, Tri-Secret Secure to Business CriticalDay 3Day 3: Editions
Match multi-cluster + 90-day Time Travel + masking to EnterpriseDay 3Day 3: Editions
Identify which interface can run PUT/GET (and which can’t)Day 4Day 4: Interfaces
List 8 of the 12 database objects without peekingDay 5Day 5: Object hierarchy
Pick ACCOUNT_USAGE vs INFORMATION_SCHEMA for a given queryDay 5Day 5: Account Usage
Resolve a 3-layer parameter precedence scenario in 10 secondsDay 6Day 6: Parameters
Distinguish session-parameter cascade from object-parameter cascadeDay 6Day 6: Parameters
🎯 Exam Tip

Read every option fully before picking. Domain 1 questions are built around two specific distractor types. One is the “technically true but not the BEST answer” option. The other is the “close but wrong on one detail” option. The qualifier words tell you which trap you are in. MINIMUM edition means start at Standard and climb only as far as the requirement forces you. Business Critical for PHI, not VPS. MOST appropriate means eliminate the over-spec answer first. ALL EXCEPT means flip the question: find the one item that does not fit. On multi-select, the question states the exact number to pick. Partial credit is rare on the COF-C03; both answers must be right.

📝 Week 1 Wrap

You’ve covered: Snowflake as cloud-native SaaS on AWS/Azure/GCP, the 3-layer architecture (Storage / Compute / Cloud Services), and the four editions with their boundary features. Plus the interfaces (Snowsight, Snowflake CLI, SnowSQL, VS Code), the 12 database objects, INFORMATION_SCHEMA vs ACCOUNT_USAGE, and the “most-specific scope wins” parameter rule.

Take the Week 1 Practice Test (10 questions, mixed domains). Target: 8/10. If you score 6 or below, re-read the days flagged in your wrong answers before starting Week 2.

Tomorrow (Day 8): Week 2 begins. We move into Domain 1 sub-objective 1.4: Virtual Warehouses: sizes, Gen 1 vs Gen 2, Snowpark-Optimized, auto-suspend, and the 60-second minimum billing. That is the first day of a six-day stretch on storage and warehouses. The stretch builds toward micro-partitions, table types, Iceberg, Dynamic Tables, and view types by Day 13.

Abhay Krishnan

Abhay Krishnan

Senior Data & AI Consultant
Connect on LinkedIn

With over five years of data engineering experience at EY and Infosys, Abhay Krishnan specializes in building scalable data pipelines and cloud warehousing solutions. He is a certified SnowPro Core professional, alongside credentials in AWS and Azure. Abhay created this 50-day track to solve a problem he faced firsthand: the lack of a structured, free resource for Snowflake certification prep. Follow him on LinkedIn for more data engineering insights.