ADA-C01 Question 224
Single answerMonitor multi-cluster warehousesA data engineering team uses a multi-cluster virtual warehouse named ETL_WH to support hundreds of concurrent ELT jobs during a daily ingestion window. The warehouse is configured with MIN_CLUSTER_COUNT = 1, MAX_CLUSTER_COUNT = 4, and SCALING_POLICY = STANDARD. Users report that during the busiest hour, some queries still spend significant time queued, but the Snowflake administrator is not sure whether the warehouse is actually scaling out when needed or whether the queueing is caused by another issue. Which approach should the administrator use to most directly monitor whether the multi-cluster warehouse is adding clusters and whether queries are waiting on warehouse resources?
- A
Query ACCOUNT_USAGE.WAREHOUSE_LOAD_HISTORY for ETL_WH to review load metrics such as queued load, and correlate that with warehouse/cluster activity from warehouse monitoring views to see whether additional clusters were used during the period.
- B
Query ACCOUNT_USAGE.QUERY_HISTORY only, because it contains all information needed to determine the exact number of clusters active for the warehouse at any point in time.
- C
Use SHOW WAREHOUSES to determine how many clusters were active during the busy hour, because it returns the historical scaling activity for each warehouse.
- D
Review DATABASE_STORAGE_USAGE_HISTORY for the ETL database to determine whether the warehouse added clusters, since increased storage activity indicates multi-cluster scaling.
Show answer and explanation
Correct answer: A
Explanation
To monitor multi-cluster warehouses effectively, administrators should focus on warehouse-specific telemetry rather than query-only, current-state, or storage-oriented views. In Snowflake, ACCOUNT_USAGE.WAREHOUSE_LOAD_HISTORY is a key source for analyzing warehouse load, including queued load, which helps identify concurrency pressure and resource contention. For a multi-cluster warehouse, this should be correlated with warehouse usage/monitoring information to determine whether Snowflake scaled the warehouse out by adding clusters during the observed interval. QUERY_HISTORY can complement this analysis by showing query-level queue times, but it is not the best single source for confirming multi-cluster scaling behavior. SHOW WAREHOUSES is configuration-oriented and current-state oriented, not historical monitoring. DATABASE_STORAGE_USAGE_HISTORY is unrelated to compute scaling. This aligns with Snowflake best practices: use Account Usage and warehouse monitoring views to assess queueing, load, and warehouse behavior over time when troubleshooting multi-cluster performance.
- A. Correct.
Correct. For monitoring a multi-cluster warehouse, the administrator should use Snowflake's warehouse monitoring/account usage views that expose warehouse load and queueing behavior. WAREHOUSE_LOAD_HISTORY is specifically useful for understanding queued load and execution pressure over time. Correlating that with warehouse usage/monitoring data is the most direct way to determine whether the warehouse scaled out and whether queries were queued due to insufficient warehouse resources.
- B. Incorrect.
Incorrect. QUERY_HISTORY is useful for examining individual query behavior, including queue-related timing fields, but it does not by itself provide the authoritative historical count of active clusters for the warehouse over time. Relying only on QUERY_HISTORY can help identify that queries queued, but not directly confirm warehouse scaling behavior as effectively as warehouse load/usage monitoring views.
- C. Incorrect.
Incorrect. SHOW WAREHOUSES returns current warehouse properties and state, not a detailed historical record of scaling activity during a prior busy hour. It can confirm that a warehouse is configured as multi-cluster, but it is not the right tool for monitoring past cluster additions or queue trends.
- D. Incorrect.
Incorrect. DATABASE_STORAGE_USAGE_HISTORY tracks storage consumption, not warehouse compute scaling. Storage growth or activity does not indicate whether a multi-cluster warehouse added clusters in response to concurrency.