Point-in-time restore for block blobs
Point-in-time restore provides protection against accidental deletion or corruption by enabling you to restore block blob data to an earlier state. Point-in-time restore is useful in scenarios where a user or application accidentally deletes data or where an application error corrupts data. Point-in-time restore also enables testing scenarios that require reverting a data set to a known state before running further tests.
Point-in-time restore is supported for general-purpose v2 storage accounts in the standard performance tier only. Only data in the hot and cool access tiers can be restored with point-in-time restore.
To learn how to enable point-in-time restore for a storage account, see Perform a point-in-time restore on block blob data.
How point-in-time restore works
To enable point-in-time restore, you create a management policy for the storage account and specify a retention period. During the retention period, you can restore block blobs from the present state to a state at a previous point in time.
To initiate a point-in-time restore, call the Restore Blob Ranges operation and specify a restore point in UTC time. You can specify lexicographical ranges of container and blob names to restore, or omit the range to restore all containers in the storage account. Up to 10 lexicographical ranges are supported per restore operation.
Azure Storage analyzes all changes that have been made to the specified blobs between the requested restore point, specified in UTC time, and the present moment. The restore operation is atomic, so it either succeeds completely in restoring all changes, or it fails. If there are any blobs that cannot be restored, then the operation fails, and read and write operations to the affected containers resume.
The following diagram shows how point-in-time restore works. One or more containers or blob ranges is restored to its state n days ago, where n is less than or equal to the retention period defined for point-in-time restore. The effect is to revert write and delete operations that happened during the retention period.
Only one restore operation can be run on a storage account at a time. A restore operation cannot be canceled once it is in progress, but a second restore operation can be performed to undo the first operation.
The Restore Blob Ranges operation returns a restore ID that uniquely identifies the operation. To check the status of a point-in-time restore, call the Get Restore Status operation with the restore ID returned from the Restore Blob Ranges operation.
When you perform a restore operation, Azure Storage blocks data operations on the blobs in the ranges being restored for the duration of the operation. Read, write, and delete operations are blocked in the primary location. For this reason, operations such as listing containers in the Azure portal may not perform as expected while the restore operation is underway.
Read operations from the secondary location may proceed during the restore operation if the storage account is geo-replicated.
Point-in-time restore supports restoring against operations that acted on block blobs only. Any operations that acted on containers cannot be restored. For example, if you delete a container from the storage account by calling the Delete Container operation, that container cannot be restored with a point-in-time restore operation. Rather than deleting an entire container, delete individual blobs if you may want to restore them later.
Prerequisites for point-in-time restore
Point-in-time restore requires that the following Azure Storage features be enabled before you can enable point-in-time restore:
To learn more about Microsoft's recommendations for data protection, see Data protection overview.
After you enable blob versioning for a storage account, every write operation to a blob in that account results in the creation of a new version. For this reason, enabling blob versioning may result in additional costs. To minimize costs, use a lifecycle management policy to automatically delete old versions. For more information about lifecycle management, see Optimize costs by automating Azure Blob Storage access tiers.
Retention period for point-in-time restore
When you enable point-in-time restore for a storage account, you specify a retention period. Block blobs in your storage account can be restored during the retention period.
The retention period begins a few minutes after you enable point-in-time restore. Keep in mind that you cannot restore blobs to a state prior to the beginning of the retention period. For example, if you enabled point-in-time restore on May 1st with a retention of 30 days, then on May 15th you can restore to a maximum of 15 days. On June 1st, you can restore data from between 1 and 30 days.
The retention period for point-in-time restore must be at least one day less than the retention period specified for soft delete. For example, if the soft delete retention period is set to 7 days, then the point-in-time restore retention period may be between 1 and 6 days.
The retention period that you specify for point-in-time restore has no effect on the retention of blob versions. Blob versions are retained until they are explicitly deleted. To optimize costs by deleting or tiering older versions, create a lifecycle management policy. For more information, see Optimize costs by automatically managing the data lifecycle.
The time that it takes to restore a set of data is based on the number of write and delete operations made during the restore period. For example, an account with one million blobs with 3,000 blobs added per day and 1,000 blobs deleted per day will require approximately two hours to restore to a point 30 days in the past. A retention period and restoration more than 90 days in the past would not be recommended for an account with this rate of change.
Permissions for point-in-time restore
To initiate a restore operation, a client must have write permissions to all containers in the storage account. To grant permissions to authorize a restore operation with Azure Active Directory (Azure AD), assign the Storage Account Contributor role to the security principal at the level of the storage account, resource group, or subscription.
Limitations and known issues
Point-in-time restore for block blobs has the following limitations and known issues:
- Only block blobs in a standard general-purpose v2 storage account can be restored as part of a point-in-time restore operation. Append blobs, page blobs, and premium block blobs are not restored.
- If you have deleted a container during the retention period, that container will not be restored with the point-in-time restore operation. If you attempt to restore a range of blobs that includes blobs in a deleted container, the point-in-time restore operation will fail. To learn about protecting containers from deletion, see Soft delete for containers.
- If you use permanent delete to purge soft-deleted versions of a blob during the point-in-time restore retention period, then a restore operation may not be able to restore that blob correctly.
- If a blob has moved between the hot and cool tiers in the period between the present moment and the restore point, the blob is restored to its previous tier. Restoring block blobs in the archive tier is not supported. For example, if a blob in the hot tier was moved to the archive tier two days ago, and a restore operation restores to a point three days ago, the blob is not restored to the hot tier. To restore an archived blob, first move it out of the archive tier. For more information, see Overview of blob rehydration from the archive tier.
- If an immutability policy is configured, then a restore operation can be initiated, but any blobs that are protected by the immutability policy will not be modified. A restore operation in this case will not result in the restoration of a consistent state to the date and time given.
- A block that has been uploaded via Put Block or Put Block from URL, but not committed via Put Block List, is not part of a blob and so is not restored as part of a restore operation.
- A blob with an active lease cannot be restored. If a blob with an active lease is included in the range of blobs to restore, the restore operation will fail atomically. Break any active leases prior to initiating the restore operation.
- Performing a customer-managed failover on a storage account resets the earliest possible restore point for that storage account. For example, suppose you have set the retention period to 30 days. If more than 30 days have elapsed since the failover, then you can restore to any point within that 30 days. However, if fewer than 30 days have elapsed since the failover, then you cannot restore to a point prior to the failover, regardless of the retention period. For example, if it's been 10 days since the failover, then the earliest possible restore point is 10 days in the past, not 30 days in the past.
- Snapshots are not created or deleted as part of a restore operation. Only the base blob is restored to its previous state.
- Point-in-time restore is not supported for hierarchical namespaces or operations via Azure Data Lake Storage Gen2.
If you restore block blobs to a point that is earlier than September 22, 2020, preview limitations for point-in-time restore will be in effect. Microsoft recommends that you choose a restore point that is equal to or later than September 22, 2020 to take advantage of the generally available point-in-time restore feature.
Support for this feature might be impacted by enabling Data Lake Storage Gen2, Network File System (NFS) 3.0 protocol, or the SSH File Transfer Protocol (SFTP).
If you've enabled any of these capabilities, see Blob Storage feature support in Azure Storage accounts to assess support for this feature.
Pricing and billing
There is no charge to enable point-in-time restore. However, enabling point-in-time restore also enables blob versioning, soft delete, and change feed, each of which may result in additional charges.
Billing for performing point-in-time restores is based on the amount of changefeed data processed for the restore. You are also billed for any storage transactions involved in the restore process.
For more information about pricing for point-in-time restore, see Block blob pricing.
Send inn og vis tilbakemelding for