ADA-C01 Question 254
Single answer3.4 Given a scenario, manage tasks.A data engineering team uses Snowflake tasks to run an hourly ELT pipeline. The root task loads raw files, and two child tasks transform and publish data after the load finishes. Recently, the team noticed duplicate processing whenever the root task runtime exceeds 60 minutes, causing the next scheduled run to start before the previous task graph has finished. The administrator must prevent overlapping executions of the task graph while keeping the existing hourly schedule. Which action should the administrator take?
- A
Set ALLOW_OVERLAPPING_EXECUTION = FALSE on the root task of the task graph
- B
Increase the USER_TASK_TIMEOUT_MS value on each task so the next scheduled run waits for the current run to finish
- C
Suspend the child tasks and recreate them as standalone scheduled tasks with a later CRON expression
- D
Configure a larger warehouse for the child tasks only, because overlapping execution is controlled at the child task level
Show answer and explanation
Correct answer: A
Explanation
In Snowflake, task graphs are managed from the root task, and graph-level concurrency behavior is controlled with the ALLOW_OVERLAPPING_EXECUTION parameter on that root task. When set to FALSE, Snowflake does not start a new run of the graph if the prior run has not completed. This is the most direct administrative control for preventing duplicate processing caused by overlapping scheduled executions. By contrast, USER_TASK_TIMEOUT_MS manages execution time limits, not schedule concurrency. Adjusting warehouse size can improve performance but is not a concurrency control. Replacing a task graph with disconnected scheduled tasks is generally not a best practice when built-in dependency orchestration already exists. This aligns with Snowflake documentation on CREATE TASK and ALTER TASK behavior for task graphs and root-task parameters.
- A. Correct.
Correct. For a task graph, overlapping runs are controlled by the root task property ALLOW_OVERLAPPING_EXECUTION. Setting it to FALSE prevents a new instance of the task graph from starting while a previous run is still in progress. This directly addresses the scenario where the hourly schedule triggers again before the current graph finishes.
- B. Incorrect.
Incorrect. USER_TASK_TIMEOUT_MS controls how long a task is allowed to run before timing out; it does not serialize scheduled runs or prevent a later schedule from attempting another execution. Increasing the timeout may allow longer execution, but it does not solve overlapping graph runs.
- C. Incorrect.
Incorrect. Rebuilding the graph as separate scheduled tasks adds operational complexity and can break dependency-based orchestration. It also does not guarantee safe sequencing in the same way as a task graph with overlapping execution disabled. Snowflake task graphs are designed specifically to manage parent-child dependencies.
- D. Incorrect.
Incorrect. A larger warehouse may reduce runtime, but it does not control whether a new scheduled graph instance can start while another instance is already active. In addition, overlapping execution for a graph is governed by the root task setting, not by child-task-level compute configuration.