What is 'Log Memory' in Query Store 2017

If we look at the documentation for the underlying object, sys.query_store_runtime_stats, we'll see it has the following descriptions:

• avg_log_bytes_used - Average number of bytes in the database log used by the query plan, within the aggregation interval.
  Note: Azure SQL Data Warehouse will always return zero (0).
• last_log_bytes_used - Number of bytes in the database log used by the last execution of the query plan, within the aggregation interval.
  Note: Azure SQL Data Warehouse will always return zero (0).
• min_log_bytes_used - Minimum number of bytes in the database log used by the query plan, within the aggregation interval.
  Note: Azure SQL Data Warehouse will always return zero (0).
• max_log_bytes_used - Maximum number of bytes in the database log used by the query plan, within the aggregation interval.
  Note: Azure SQL Data Warehouse will always return zero (0).
• stdev_log_bytes_used - Standard deviation of the number of bytes in the database log used by a query plan, within the aggregation interval.
  Note: Azure SQL Data Warehouse will always return zero (0).

LowlyDBA's answer covers what the metrics actually mean. This answer is just to explain why the numbers in the Query Store user interface don't totally make sense.

Getting Some Log Data

First, let's to get data in those columns on SQL Server 2017 Developer Edition on my laptop.

Create a database:

USE [master];
GO

CREATE DATABASE [231682];
GO

Enable Query Store with very impractical settings:

ALTER DATABASE [231682] SET QUERY_STORE = ON (INTERVAL_LENGTH_MINUTES = 1);

Do something that will generate some transaction log usage:

USE [231682];

CREATE TABLE dbo.Junk
(
    Id INT NOT NULL,
    MoreJunk NVARCHAR(MAX) NOT NULL
);

INSERT INTO dbo.Junk
    (Id, MoreJunk)
SELECT TOP 1000
    m.message_id, m.[text]
FROM sys.messages m;

Force flush to disk in case it hasn't happened yet:

EXEC sp_query_store_flush_db;

Voila:

SELECT 
    qsrs.avg_log_bytes_used, 
    qsrs.last_log_bytes_used, 
    qsrs.min_log_bytes_used, 
    qsrs.max_log_bytes_used, 
    qsrs.stdev_log_bytes_used
FROM sys.query_store_runtime_stats qsrs
WHERE qsrs.avg_log_bytes_used > 0;

Calculation Problem

From the query store user interface standpoint, the calculation being run looks like this:

SELECT TOP (@results_row_count)
    -- other columns
    ROUND(CONVERT(float, SUM(rs.avg_log_bytes_used*rs.count_executions))*1024,2) total_log_bytes_used,
    -- other columns
FROM sys.query_store_runtime_stats rs
    JOIN sys.query_store_plan p ON p.plan_id = rs.plan_id
    JOIN sys.query_store_query q ON q.query_id = p.query_id
    JOIN sys.query_store_query_text qt ON q.query_text_id = qt.query_text_id
WHERE NOT (rs.first_execution_time > @interval_end_time OR rs.last_execution_time < @interval_start_time)
GROUP BY p.query_id, qt.query_sql_text, q.object_id
HAVING COUNT(distinct p.plan_id) >= 1
ORDER BY total_log_bytes_used DESC

There is a bug in the calculation, it should be dividing by 1,024 to go from bytes to kilobytes. As it stands, it's multiplying the bytes by 1,024, and then reporting them as kilobytes - which makes them appear to be off by a factor of ~1,000,000.

For example, my repro here produced 346,796 bytes of log over 1 execution of the query. The Query Store user interface, instead of showing 338 kilobytes, shows 355,119,104 kilobytes instead.

I've reported this problem to Microsoft: Query Store "Log Memory Used" metric calculation is wrong