# Free ARA-C01 practice questions

> 10 free ARA-C01 practice questions with the correct answer and an explanation.

Page: https://devopscodelab.com/exams/snowpro-advanced-architect/free-practice-questions/
Practice exams: https://devopscodelab.com/exams/snowpro-advanced-architect/

### Question 1 of 10

A service account used by an ETL tool must authenticate without a password and without interactive MFA prompts, and the security team wants to rotate credentials with no downtime. Which approach should the architect recommend?

- A. Password authentication with a password policy requiring rotation every 30 days and a minimum length of 14 characters
- B. SAML single sign-on through the corporate IdP with an exception that skips MFA for the ETL service account
- C. Key-pair authentication, setting RSA_PUBLIC_KEY_2 to the new key before removing the old RSA_PUBLIC_KEY value
- D. Snowflake OAuth with a refresh token validity set to the maximum so that tokens rarely need to be reissued

**Answer:** C

**Explanation:** Key-pair authentication is designed for programmatic clients; Snowflake supports two public key slots per user, so a new key can be registered in RSA_PUBLIC_KEY_2, clients switched over, and the old key removed with no downtime. Passwords for service users are discouraged and rotating them forces coordinated cut-overs. SAML SSO requires a browser-based flow unsuitable for unattended ETL. Long-lived OAuth refresh tokens still require an initial interactive consent and extending them weakens security rather than enabling clean rotation.

### Question 2 of 10

A data platform team wants to grant SELECT on every table that will ever be created in the SALES.RAW schema to the ANALYST role, but table owners must not be able to grant access to their tables themselves. Which design meets both requirements?

- A. Define future grants at the database level on SALES and grant MANAGE GRANTS to the role that loads data into RAW
- B. Keep SALES.RAW as a regular schema and have each object owner run GRANT SELECT after creating a new table
- C. Grant OWNERSHIP of the SALES.RAW schema to ANALYST so that the role automatically sees every table created in it
- D. Create SALES.RAW as a managed access schema and define future grants of SELECT on tables in that schema to ANALYST

**Answer:** D

**Explanation:** In a managed access schema only the schema owner or a role with MANAGE GRANTS can grant privileges on contained objects, so table owners lose discretionary grant control, and future grants automatically give ANALYST SELECT on new tables. Leaving the schema regular lets object owners grant access themselves, which violates the requirement. Giving ANALYST ownership of the schema grants far more than read access. Granting MANAGE GRANTS to the loading role gives it account-wide grant power, the opposite of what is needed.

### Question 3 of 10

A company wants all connections from its corporate offices to be allowed while all other client IP addresses are blocked, and it wants to manage the address list as a reusable schema-level object that several policies can reference. Which approach should be used?

- A. Create a session policy with a short idle timeout and attach it to the account so that unknown clients are disconnected
- B. Create a security integration of type SAML2 so that only users federated through the corporate IdP can sign in
- C. Create network rules for the office CIDR ranges and list them in a network policy's ALLOWED_NETWORK_RULE_LIST
- D. Create an authentication policy that restricts CLIENT_TYPES to Snowflake UI and drivers used on the corporate network

**Answer:** C

**Explanation:** Network rules are schema-level objects that group IP ranges (or other identifiers) and can be referenced by one or more network policies through ALLOWED_NETWORK_RULE_LIST, which is the current recommended way to restrict client IPs. A session policy controls idle session timeouts, not source addresses. Authentication policies control authentication methods and client types, not IP ranges. A SAML integration controls who can authenticate but does not block traffic from outside the corporate network.

### Question 4 of 10

A consumer in a different cloud region wants a provider's data product. The provider does not want to set up and manage replication manually for every consumer region. Which approach should the provider use?

- A. Publish a listing with Cross-Cloud Auto-Fulfillment so Snowflake replicates the data to consumer regions
- B. Create a reader account in the provider's region and give the consumer credentials to query the data there
- C. Create a direct share and add the consumer's account locator so that the share works across regions automatically
- D. Unload the data to an external stage each night and let the consumer create an external table in its own region

**Answer:** A

**Explanation:** Listings with Cross-Cloud Auto-Fulfillment let Snowflake replicate the data product to other regions and clouds on demand, removing the manual replication work. A direct share only works within the same region; cross-region sharing otherwise requires replicating the database first. A reader account in the provider region works but forces the consumer to use a separate, provider-managed account rather than their own. Nightly unloading adds a pipeline, latency and egress management, which the provider wanted to avoid.

### Question 5 of 10

A software vendor wants to distribute an application to Snowflake customers that includes stored procedures, a Streamlit UI and proprietary reference data. Customers must run it in their own accounts, and the vendor's code must not be visible to them. Which approach should the vendor use?

- A. Deploy it with Snowpark Container Services in the vendor's account and expose a public endpoint to customers
- B. Publish the code to a Git repository and give customers a script that creates the objects in their account
- C. Package it as a Snowflake Native App and distribute it through a listing on Snowflake Marketplace
- D. Share a database containing secure UDFs and the reference data through a direct share to each customer

**Answer:** C

**Explanation:** The Native App Framework packages code, Streamlit UIs and data into an application that consumers install in their own accounts from a listing, while protecting the provider's logic. A direct share can expose secure UDFs and data but cannot ship stored procedures and a Streamlit UI as an installable application. Handing over scripts exposes all source code. Running the app in the vendor's account means it does not run in the customers' accounts, contrary to the requirement.

### Question 6 of 10

A company runs its production account in AWS us-east-1 and needs a disaster recovery plan in which a secondary account in us-west-2 can take over databases, roles, users and warehouses, and BI tools reconnect without changing connection strings. Which combination should the architect design?

- A. Database replication of the production databases plus a CNAME record that operators update during an outage
- B. A replication group containing the databases plus a secure share from the primary to the secondary account
- C. A failover group containing only the databases plus Snowpipe loading the same files into both accounts
- D. A failover group containing the databases and account objects plus a connection object for client redirect

**Answer:** D

**Explanation:** Failover groups replicate databases and account-level objects such as users, roles and warehouses and can be promoted in the secondary account, while a connection object provides a URL that can be redirected so clients reconnect without changes. Replication groups cannot be failed over and shares do not replicate account objects. Manually editing DNS is error-prone and plain database replication omits users and roles. Loading data twice through Snowpipe duplicates pipelines and still omits account objects from failover.

### Question 7 of 10

A transformation pipeline must keep a summary table no more than 10 minutes behind its source tables. The team wants to define the result declaratively with a SELECT statement and let Snowflake work out refresh scheduling. Which object should be used?

- A. A dynamic table with TARGET_LAG = '10 minutes' defined on the SELECT statement
- B. A stream on each source table and a task that runs a MERGE every 10 minutes
- C. A materialized view defined on the SELECT statement, joining all of the source tables
- D. A secure view defined on the SELECT statement, cached by the result cache for 10 minutes

**Answer:** A

**Explanation:** Dynamic tables are declared with a query and a target lag, and Snowflake schedules incremental or full refreshes to meet that lag. Materialized views cannot contain joins, so a multi-table summary is not possible. Streams and tasks work but require writing and maintaining imperative MERGE logic, which the team wants to avoid. A secure view computes results at query time and the result cache is invalidated when data changes, so it does not provide a maintained summary table.

### Question 8 of 10

An IoT platform sends millions of small rows per second from a Java application and requires that rows be queryable within seconds without first writing files to a stage. Which ingestion method should the architect choose?

- A. Snowpipe Streaming using the Snowpipe Streaming SDK to write rows directly into the target tables
- B. Hybrid tables accepting single-row INSERT statements from many concurrent JDBC connections
- C. Snowpipe with auto-ingest, with the application writing one small file per second to the external stage
- D. A COPY INTO task scheduled every 10 seconds on a multi-cluster warehouse in auto-scale mode

**Answer:** A

**Explanation:** Snowpipe Streaming ingests rows directly through the SDK or Kafka connector without staging files, giving low latency and lower cost for high-volume row streams. Writing tiny files for file-based Snowpipe adds per-file overhead and latency. Tasks cannot run more frequently than the minimum schedule and would keep a warehouse running continuously. Hybrid tables target transactional workloads with point lookups, not analytical ingestion of millions of rows per second.

### Question 9 of 10

A finance team wants a hard cap on warehouse credits for a department's warehouses: at 90% of the monthly quota users should be notified, and at 100% running queries must be cancelled. Which configuration achieves this?

- A. A resource monitor on those warehouses with a NOTIFY trigger at 90% and SUSPEND_IMMEDIATE at 100%
- B. A budget on the department's warehouses with a 90% notification and a stored procedure that drops the warehouses at 100%
- C. A warehouse-level STATEMENT_TIMEOUT_IN_SECONDS value calculated so that the monthly credit quota cannot be exceeded
- D. A resource monitor with a credit quota, a NOTIFY trigger at 90% and a SUSPEND trigger at 100%, assigned at the account level

**Answer:** A

**Explanation:** A resource monitor assigned to the department's warehouses can notify at 90% and, with SUSPEND_IMMEDIATE at 100%, suspend the warehouses and cancel running queries. The SUSPEND action lets running queries finish, and an account-level monitor would affect every warehouse. Budgets provide notifications but dropping warehouses with a procedure is destructive and not a quota control. A statement timeout limits individual queries, not cumulative monthly credits.

### Question 10 of 10

Users run highly selective queries such as WHERE EMAIL = 'x@y.com' against a 5-billion-row table, and each returns a handful of rows after scanning most of the table. Which feature is designed for this pattern?

- A. A clustering key on a high-cardinality hash of EMAIL with Automatic Clustering
- B. Query acceleration service enabled on the warehouse with an unlimited scale factor
- C. A multi-cluster warehouse with a maximum of ten clusters in auto-scale mode
- D. Search optimization service on the EMAIL column using an EQUALITY search method

**Answer:** D

**Explanation:** Search optimization builds a persistent search access path that drastically improves selective point-lookup queries on large tables. Clustering on a high-cardinality hash is expensive to maintain and gives little pruning benefit for equality lookups. Multi-cluster warehouses address concurrency, not the efficiency of a single query. Query acceleration offloads scanning to serverless compute but still scans the data, unlike search optimization's pruning.

Ready for a timed run? Take the free 15-question ARA-C01 mini exam: https://devopscodelab.com/app/exams/snowpro-advanced-architect/mini/

DevOpsCode Lab is not affiliated with Snowflake. ARA-C01 and related names are trademarks of their owners. We write every question ourselves; these are not real exam questions.
