A company built data pipelines using dynamic tables in Snowflake. Recently, some dynamic table refreshes have been taking longer than expected, causing delays in data availability for downstream…

Question

Options

• AIncrease the refresh table frequency to keep the data fresh and improve reporting accuracy.
• BUse Snowsight or the DYNAMIC_TABLE_REFRESH_HISTORY function to monitor the refresh
• CSet the dynamic tables to full refresh mode to simplify the monitoring process and ensure
• DMonitor the QUERY_HISTORY view to analyze the performance of the virtual warehouse used to
• ESchedule this query to analyze the dynamic table performance:

• A
  3% (1)
• B
  80% (24)
• C
  7% (2)
• E
  10% (3)

Explanation

Monitoring inefficient dynamic table refreshes requires purpose-built observability tools: Snowsight and DYNAMIC_TABLE_REFRESH_HISTORY (B) provide direct visibility into refresh durations, lag, and failure states specific to dynamic tables, while QUERY_HISTORY (D) lets you drill into the underlying warehouse queries driving those refreshes - revealing slow SQL, spilled data, and resource bottlenecks that explain cost increases.

Why the distractors are wrong:

• A is counterproductive - increasing refresh frequency would worsen the performance and cost problem, not fix it; you'd refresh more often without fixing the root cause.
• C switching to full refresh mode makes performance worse for large tables (it reprocesses everything instead of only changed data) and adds zero monitoring capability.
• E is incomplete as shown - it implies a custom scheduled query, but the built-in system views (B and D) already provide this data without manual scheduling overhead.

Memory tip: Think "two layers of monitoring" - the dynamic table layer (B: refresh history, lag, errors) and the warehouse layer (D: query cost, execution time). Both are needed because a refresh can fail at either layer. The keyword to trigger this is "delay + cost" → you need both refresh metadata and query execution data to diagnose the problem.

No community discussion yet for this question.