Query history and profiling in Snowflake - Commands & Configuration

Introduction

When you run queries in Snowflake, you may want to see what queries ran before and how long they took. Query history and profiling help you find slow queries and understand their details to improve performance.

When you want to check which queries used the most time or resources in your Snowflake account.

When you need to find errors or failed queries to fix issues quickly.

When you want to see who ran a query and when for auditing purposes.

When you want to analyze query performance to optimize your database usage.

When you want to track query trends over time to plan capacity.

Commands

This command fetches the last 5 queries run in the past day with details like who ran them, when, how long they took, and if they had errors. It helps you quickly see recent query activity.

Terminal

SELECT query_id, user_name, start_time, end_time, total_elapsed_time, query_text, error_message FROM snowflake.account_usage.query_history WHERE start_time > DATEADD(day, -1, CURRENT_TIMESTAMP()) ORDER BY start_time DESC LIMIT 5;

Expected Output

QUERY_ID                             USER_NAME   START_TIME                  END_TIME                    TOTAL_ELAPSED_TIME   QUERY_TEXT                             ERROR_MESSAGE

01a2b3c4-d5e6-789f-0123-456789abcdef   JOHN_DOE   2024-06-01 10:15:00.000   2024-06-01 10:15:05.000   5000                SELECT * FROM sales WHERE region = 'US'   

02b3c4d5-e6f7-890a-1234-56789abcdef0   JANE_SMITH 2024-06-01 09:50:00.000   2024-06-01 09:50:02.000   2000                UPDATE customers SET status = 'active' WHERE last_login > '2024-01-01'   

03c4d5e6-f7a8-901b-2345-6789abcdef01   JOHN_DOE   2024-06-01 09:30:00.000   2024-06-01 09:30:10.000   10000               SELECT COUNT(*) FROM orders   

04d5e6f7-a8b9-012c-3456-789abcdef012   ALICE_W   2024-06-01 09:00:00.000   2024-06-01 09:00:03.000   3000                DELETE FROM temp_data WHERE created_at < '2024-01-01'   

05e6f7a8-b9c0-123d-4567-89abcdef0123   BOB_K     2024-06-01 08:45:00.000   2024-06-01 08:45:01.500   1500                SELECT * FROM products WHERE price > 100   

This command shows detailed timing for a specific query by its ID, breaking down total time into compilation and execution. It helps you understand where time is spent.

Terminal

SELECT query_id, query_text, total_elapsed_time, compilation_time, execution_time FROM snowflake.account_usage.query_history WHERE query_id = '01a2b3c4-d5e6-789f-0123-456789abcdef';

Expected Output

QUERY_ID                             QUERY_TEXT                             TOTAL_ELAPSED_TIME   COMPILATION_TIME   EXECUTION_TIME

01a2b3c4-d5e6-789f-0123-456789abcdef   SELECT * FROM sales WHERE region = 'US'   5000                1000               4000

This command retrieves the query profile for the given query ID. It shows detailed steps and resource usage during query execution to help diagnose performance issues.

Terminal

SELECT * FROM table(information_schema.query_profile('01a2b3c4-d5e6-789f-0123-456789abcdef'));

Expected Output

OPERATION          START_TIME           END_TIME             ROWS_PRODUCED   BYTES_PRODUCED

Scan               2024-06-01 10:15:00 2024-06-01 10:15:02  10000           800000

Filter             2024-06-01 10:15:02 2024-06-01 10:15:04  5000            400000

Result             2024-06-01 10:15:04 2024-06-01 10:15:05  5000            400000

Key Concept

If you remember nothing else from this pattern, remember: query history shows what ran and when, and query profiling shows how the query used resources and time.

Common Mistakes

Trying to query query_history without specifying a recent time range.

The query_history table can be very large and slow to query without filters, causing timeouts or delays.

Always filter query_history by a recent time window like the last day or week.

Using query_profile without a valid query_id or for queries older than retention period.

Query profiles are only available for recent queries and require a correct query ID, otherwise the command returns no data.

Use a valid recent query_id from query_history to get the profile.

Summary

Use SELECT on snowflake.account_usage.query_history to see recent queries and their details.

Filter query_history by time to get fast and relevant results.

Use information_schema.query_profile with a query ID to see detailed execution steps and timings.

Practice

1. What is the main purpose of the QUERY_HISTORY view in Snowflake?

easy

A. To see details of past queries executed in the system

B. To create new tables and schemas

C. To manage user permissions and roles

D. To monitor network traffic between Snowflake and clients

Solution

1. Step 1: Understand the role of QUERY_HISTORY

   The QUERY_HISTORY view stores information about queries that have already run, including their text, execution time, and status. Its main purpose is to see details of past queries executed in the system.

2. Final Answer:

   To see details of past queries executed in the system -> Option A

3. Quick Check:

   QUERY_HISTORY = past query details [OK]

Hint: QUERY_HISTORY shows past queries and their info [OK]

Common Mistakes:

• Confusing QUERY_HISTORY with user management
• Thinking it manages network or security settings
• Assuming it creates or modifies database objects

2. Which SQL clause correctly filters queries executed by a specific user in the QUERY_HISTORY view?

easy

A. FILTER BY USER = 'john_doe'

B. WHERE USER_NAME = 'john_doe'

C. SELECT USER_NAME FROM QUERY_HISTORY WHERE 'john_doe'

D. HAVING USER_NAME = 'john_doe'

Solution

1. Step 1: Recall SQL filtering syntax

   To filter rows in SQL, the WHERE clause is used with a condition like USER_NAME = 'value'. WHERE USER_NAME = 'john_doe' is valid and standard SQL syntax.

2. Final Answer:

   WHERE USER_NAME = 'john_doe' -> Option B

3. Quick Check:

   Filter with WHERE clause = WHERE USER_NAME = 'john_doe' [OK]

Hint: Use WHERE to filter rows by user name [OK]

Common Mistakes:

• Using FILTER BY instead of WHERE
• Misplacing HAVING without GROUP BY
• Incorrect SELECT syntax without WHERE

3. Given the query:

SELECT query_text, total_elapsed_time FROM TABLE(INFORMATION_SCHEMA.QUERY_HISTORY()) WHERE execution_status = 'FAILED' ORDER BY start_time DESC LIMIT 1;

What does this query return?

medium

A. The most recent failed query's text and its total elapsed time

B. All successful queries ordered by start time

C. The oldest failed query's text and elapsed time

D. An error because QUERY_HISTORY is not a table

Solution

1. Step 1: Analyze the query clauses

   The query uses TABLE(INFORMATION_SCHEMA.QUERY_HISTORY()) to access query history, filters for execution_status = 'FAILED', orders by start_time DESC (most recent first), and limits to 1 row, returning the most recent failed query's text and total elapsed time.

2. Final Answer:

   The most recent failed query's text and its total elapsed time -> Option A

3. Quick Check:

   Filter failed + order desc + limit 1 = most recent failed query [OK]

Hint: ORDER BY DESC + LIMIT 1 gets latest record [OK]

Common Mistakes:

• Thinking QUERY_HISTORY is a normal table
• Confusing oldest vs most recent due to ORDER BY
• Ignoring the WHERE filter on execution_status

4. You wrote this query to find slow queries but it returns no results:

SELECT query_text FROM TABLE(INFORMATION_SCHEMA.QUERY_HISTORY()) WHERE total_elapsed_time > 1000;

What is the likely issue?

medium

A. The TABLE() function cannot be used with QUERY_HISTORY

B. QUERY_HISTORY does not have total_elapsed_time column

C. The query_text column cannot be selected from QUERY_HISTORY

D. total_elapsed_time is in microseconds, so 1000 is too small a threshold

Solution

1. Step 1: Check the unit of total_elapsed_time

   In Snowflake, total_elapsed_time is measured in microseconds, so 1000 microseconds (1 millisecond) is too small a threshold, and few or no queries exceed it, resulting in no results.

2. Final Answer:

   total_elapsed_time is in microseconds, so 1000 is too small a threshold -> Option D

3. Quick Check:

   Elapsed time unit = microseconds, threshold too low [OK]

Hint: Check units: elapsed time is microseconds, not milliseconds [OK]

Common Mistakes:

• Assuming elapsed time is in seconds or milliseconds
• Thinking QUERY_HISTORY lacks columns
• Misusing TABLE() function syntax

5. You want to profile query performance by grouping queries by user and calculating average execution time. Which query correctly achieves this?

hard

A. SELECT user_name, SUM(total_elapsed_time) FROM TABLE(INFORMATION_SCHEMA.QUERY_HISTORY()) WHERE execution_status = 'SUCCESS';

B. SELECT user_name, total_elapsed_time FROM QUERY_HISTORY GROUP BY user_name;

C. SELECT user_name, AVG(total_elapsed_time) AS avg_time FROM TABLE(INFORMATION_SCHEMA.QUERY_HISTORY()) GROUP BY user_name ORDER BY avg_time DESC;

D. SELECT user_name, AVG(total_elapsed_time) FROM QUERY_HISTORY WHERE total_elapsed_time > 1000 ORDER BY user_name;

Solution

1. Step 1: Identify correct aggregation and grouping

   To get average execution time per user, use AVG(total_elapsed_time) with GROUP BY user_name from TABLE(INFORMATION_SCHEMA.QUERY_HISTORY()), as in SELECT user_name, AVG(total_elapsed_time) AS avg_time ... GROUP BY user_name ORDER BY avg_time DESC.

2. Final Answer:

   SELECT user_name, AVG(total_elapsed_time) AS avg_time FROM TABLE(INFORMATION_SCHEMA.QUERY_HISTORY()) GROUP BY user_name ORDER BY avg_time DESC; -> Option C

3. Quick Check:

   Group by user + AVG + ORDER BY avg_time = SELECT user_name, AVG(total_elapsed_time) AS avg_time FROM TABLE(INFORMATION_SCHEMA.QUERY_HISTORY()) GROUP BY user_name ORDER BY avg_time DESC; [OK]

Hint: Use GROUP BY user_name with AVG for profiling [OK]

Common Mistakes:

• Missing GROUP BY when using aggregation
• Selecting columns without aggregation
• Not using TABLE() function for QUERY_HISTORY