Snowflake - Task

A task in Snowflake is a database object that lets you run SQL statements (often INSERT, COPY INTO, or CALL to a stored procedure) on a schedule or based on dependencies.

• Trigger:
  • Time-based — uses CRON expressions or simple intervals using SCHEDULE syntax (e.g., every 5 minutes).
  • Dependency-based — runs only after another task completes using AFTER syntax, which generate a task graph
• Common uses:
  • Periodic COPY INTO from a stage to a table.
  • Data transformations in ELT pipelines.
  • Chaining tasks for multi-step workflows.

Task creation workflow overview¶

1. Create a task administrator role that can run the commands in the following steps.
2. Define a new task using CREATE TASK.
   • Assign compute resources
   • Define schedules or triggers
   • Define what happens when a task fails
   • Define additional session parameters
3. Manually test tasks using EXECUTE TASK.
4. Allow the task to run continuously using ALTER TASK … RESUME.
5. Monitor task costs
6. Refine the task as needed using ALTER TASK. Ref: https://docs.snowflake.com/en/user-guide/tasks-intro

Compared with Snowflake - Dynamic table

Feature / Aspect	Dynamic Table	Dependency-based Task
Primary purpose	Keep a table automatically updated from a query	Run SQL or stored procedures in a defined sequence
Trigger	Target lag interval (Snowflake manages refresh)	Explicit AFTER dependency or schedule
Execution control	Snowflake decides when to refresh to meet target lag	Full control over exact run order and timing
Workflow complexity	Single-step, SQL-only transformation	Multi-step workflows with branching and procedural logic
Use with external systems	Not supported	Supported (e.g., send notifications, call APIs)
Non-SQL operations	Not possible	Possible via stored procedures or external functions
Latency control	Approximate — Snowflake aims to meet target lag	Exact — runs immediately after dependency finishes
Maintenance	Minimal — Snowflake handles refresh logic	You manage scheduling, chaining, and enabling tasks
Best for	Continuous, incremental data sync	ETL/ELT pipelines, staged loads, complex orchestration