If you are using Hyperscale in Azure SQL Database, you know that it is a powerful tier that lets you rapidly scale up and scale out your database according to your needs along with autoscaling storage. However, there could be situations where storage is scaled up automatically and then due to some business needs, a significant amount of data is removed/purged and a lot of free space is left within the database
Today, we are pleased to announce that database and data file shrink is available in the Hyperscale tier in preview. Now you can reduce the allocated size of a Hyperscale database using the same DBCC SHRINK* commands that you might be familiar with. This allows you to reduce the size of the databases and free up unused space to save storage costs.
How to use shrink in Hyperscale?
Using shrink is easy and straightforward. You use the same set of commands which you might have used in other tiers of Azure SQL Database or in SQL Server.
First, identify a Hyperscale database with substantial allocated but unused storage space. For definitions of allocated and used storage space, see Azure SQL Database file space management - Azure SQL Database | Microsoft Docs. Azure portal also provides this information. You can also capture the current used, allocated, and unused space in each database file by executing the following query in the database.
DECLARE @NumPagesPerGB float = 128 * 1024;
SELECT file_id AS FileId
, size / @NumPagesPerGB AS AllocatedSpaceGB
, ROUND(CAST(FILEPROPERTY(name, 'SpaceUsed') AS float)/@NumPagesPerGB,3) AS UsedSpaceGB
, ROUND((size-CAST(FILEPROPERTY(name, 'SpaceUsed') AS float))/@NumPagesPerGB,3) AS FreeSpaceGB
, ROUND(max_size / @NumPagesPerGB,3) AS MaxSizeGB
, ROUND(CAST(size - FILEPROPERTY(name, 'SpaceUsed') AS float)*100/size,3) AS UnusedSpacePercent
FROM sys.database_files
WHERE type_desc = 'ROWS'
ORDER BY file_id
A shrink operation can be initiated using either command to shrink the entire database, or DBCC SHRINKFILE command for individual data files. We recommend using DBCC SHRINKFILE, because you can run it in parallel on multiple sessions, targeting different data files. DBCC SHRINKDATABASE is a simpler choice because you only need to run one command, but it will shrink one file at a time, which can be time-consuming for larger databases.
If shrink operation fails with any error or canceled, the progress it has made so far is retained, and the same shrink command can be simply executed again to continue.
Once shrink for all data files has completed, rerun the earlier query (or check in the Azure portal) to determine the resulting reduction in the allocated storage size. If there is still a large difference between used space and allocated space, you can rebuild indexes to reduce the total number of used data pages. This may temporarily increase allocated space further, however shrinking files again after rebuilding indexes should result in a higher reduction in allocated space.
For more details about Azure SQL Database space management, see the following documentation article: Database file space management - Azure SQL Database | Microsoft Learn
Known behaviors / limitations
- Database shrink is a long-running operation. For larger databases, it may span multiple days. To avoid shrink getting interrupted, we recommend using a client that is unlikely to get disconnected from the database.
- While shrink is running, used and allocated space for the database in the Azure portal might not be reported.
- Running SHRINKFILE with a target size slightly higher than the used space in the file tends to have a higher success rate compared to setting it to the exact used space. For instance, if a file is 128 GB in total size with 50 GB used and 78 GB free, setting the target size to 55 GB results in a better space reduction compared to using 50 GB.
- When executing DBCC SHRINKFILE concurrently on multiple files, you may encounter occasional blocking between the sessions. This is expected and does not impact the outcome of shrink.
- Shrinking of the transaction log file in the Hyperscale tier is not required as it does not contribute to the allocated data size and cost. Executing DBCC SHRINKFILE(2) has no effect on the transaction log size.
Shrink is currently in preview mode and has the following limitations:
- Shrink is not allowed on unencrypted databases. Any such attempt raises the following error:
Msg 49532, Level 16, State 1, Line 1
DBCC SHRINKFILE for data files is not supported in a Hyperscale database when the database is not encrypted. Enable transparent data encryption and try again.
To find the encryption state of the database, execute the following query:
SELECT db_name(database_id) AS 'database_name'
,encryption_state_desc
FROM sys.dm_database_encryption_keys
WHERE database_id = db_id()
If encryption state is other than ENCRYPTED then shrink will not start.
Conclusion
We hope that you will find shrink useful and beneficial for your Hyperscale databases. We welcome your feedback and suggestions on how to improve it. You can contact us by adding to this blog post and we’ll be happy to get back to you. Alternatively, you can also email us at sqlhsfeedback AT microsoft DOT com. We are eager to hear from you all!
Microsoft
