ADA-C01 Question 258
Single answerManage and schedule tasksA data engineering team has a nightly Snowflake task graph that starts with a root task scheduled at 01:00 UTC. The root task uses a dedicated virtual warehouse. Several child tasks use AFTER dependencies to transform and publish data. Recently, the team noticed that when the upstream load runs longer than expected, the 01:00 root task execution can still be running when the next scheduled time arrives, causing overlapping runs and duplicate downstream processing. The administrator must prevent a new run of the task graph from starting if the previous graph run has not completed, while keeping the existing schedule and dependency structure. What is the best solution?
- A
Set the root task parameter ALLOW_OVERLAPPING_EXECUTION = FALSE so a new graph run will not start while the previous run is still in progress
- B
Suspend all child tasks and add individual schedules to each child task so they only run after the root task finishes
- C
Replace the task graph with a stream-triggered task, because triggered tasks cannot overlap with scheduled tasks
- D
Set USER_TASK_TIMEOUT_MS = 0 on the root task so the next scheduled run waits indefinitely for the current run to finish
Show answer and explanation
Correct answer: A
Explanation
The best answer is to configure the root task with ALLOW_OVERLAPPING_EXECUTION = FALSE. In Snowflake, this parameter is specifically used on the root task of a task graph to control whether multiple instances of the graph can run at the same time. When set to FALSE, Snowflake skips starting a new graph run if the previous one is still executing, which is exactly what is needed in this scenario. This is the most direct and operationally sound solution because it preserves the existing CRON/interval schedule on the root task and the AFTER dependencies on the child tasks.
The key misconception behind the other options is confusing graph orchestration with independent task scheduling or timeout management. Child tasks in a graph should generally not have their own schedules; they should run based on predecessor completion. Likewise, USER_TASK_TIMEOUT_MS is about execution timeout, not overlap prevention. Stream-triggered tasks are valuable for event-driven designs, but they are not the appropriate fix when the requirement is to retain a nightly scheduled graph.
This aligns with Snowflake task administration best practices and the documented task graph behavior for overlapping execution control on root tasks.
- A. Correct.
Correct. For task graphs, the root task supports the ALLOW_OVERLAPPING_EXECUTION parameter. Setting it to FALSE prevents a new instance of the graph from starting if the previous graph run has not finished. This directly addresses the requirement to keep the current schedule and AFTER-based dependency structure while avoiding overlapping graph executions and duplicate downstream processing.
- B. Incorrect.
Incorrect. Child tasks in a task graph are intended to be triggered by predecessor completion using AFTER, not by separate schedules. Adding schedules to child tasks breaks the graph design and can introduce independent execution timing, race conditions, and operational complexity. It also does not properly solve graph-level overlap prevention.
- C. Incorrect.
Incorrect. Stream-triggered tasks are used to run when data changes are detected, not as a general replacement for a scheduled task graph. This option changes the execution model rather than preserving the existing schedule. Also, it does not address the specific control designed for overlapping executions of scheduled task graphs.
- D. Incorrect.
Incorrect. USER_TASK_TIMEOUT_MS controls how long a task run can execute before timing out; it does not make the next scheduled run wait for the current run to finish. Setting it to 0 does not provide overlap control for task graphs and could reflect a misunderstanding of timeout behavior versus scheduling behavior.