How to Schedule Jobs in Snowflake

What is Snowflake job scheduling?

Snowflake schedules jobs with CREATE TASK. A task stores a SQL statement and a cron or interval schedule. Once enabled, Snowflake executes the task automatically without external orchestration.

How does CREATE TASK schedule jobs?

CREATE TASK defines timing through a CRON or USING WAREHOUSE clause. You can trigger tasks by time or by dependency on another task, forming DAG-style pipelines.

What are the key parameters of CREATE TASK?

SCHEDULE or CRON sets run time; WAREHOUSE specifies compute; USER_TASK_MANAGED_INITIAL_WAREHOUSE_SIZE reserves capacity; AFTER chains tasks; WHEN prevents runs unless a boolean expression is true.

How to create a daily order summary task?

Use our ecommerce schema. The task aggregates yesterday’s orders into a reporting table. Enable the task to start the schedule immediately.

CREATE OR REPLACE TASK daily_order_summary
WAREHOUSE = analytics_wh
SCHEDULE = 'USING CRON 0 3 * * * America/Los_Angeles'
AS
INSERT INTO order_metrics (summary_date, total_orders, total_revenue)
SELECT CURRENT_DATE - 1,
COUNT(*),
SUM(total_amount)
FROM Orders
WHERE order_date = CURRENT_DATE - 1;

ALTER TASK daily_order_summary RESUME;


How to chain tasks for complex workflows?

Create downstream tasks with AFTER. Snowflake guarantees that a child task starts only after its parent finishes successfully, simplifying ETL pipelines.

CREATE OR REPLACE TASK refresh_product_sales
WAREHOUSE = analytics_wh
AFTER daily_order_summary
AS
MERGE INTO product_sales USING (
SELECT product_id, SUM(quantity) AS sold
FROM OrderItems
GROUP BY product_id
) src ON Products.id = src.product_id
WHEN MATCHED THEN UPDATE SET stock = stock - src.sold;

ALTER TASK refresh_product_sales RESUME;


How to monitor and troubleshoot tasks?

Query ACCOUNT_USAGE.TASK_HISTORY or INFORMATION_SCHEMA.TASK_HISTORY. Look at LAST_FAILED_RUN_TIME and ERROR_MESSAGE to diagnose problems quickly.

What are best practices for Snowflake job scheduling?

Use dedicated, appropriately sized warehouses. Keep SQL logic idempotent. Store results in separate tables to avoid locking source data. Pause tasks during backfills to prevent double processing.

Common mistakes and fixes

Mis-sized warehouse: A warehouse too small causes long runtimes and task overlap. Fix by setting USER_TASK_MANAGED_INITIAL_WAREHOUSE_SIZE or choosing a larger warehouse.

Forgot to RESUME: CREATE TASK defaults to SUSPENDED. Enable with ALTER TASK ... RESUME, otherwise the job never runs.

FAQ

Can I schedule sub-minute tasks?

No. The shortest interval is 1 minute. Use external orchestration for faster cadence.

How do I stop a task temporarily?

Run ALTER TASK task_name SUSPEND; to pause without dropping.

Does task failure retry automatically?

Snowflake retries once after a system error but not on SQL errors. Build idempotency and error handling into your SQL.