On January 29, 2019, Microsoft was disrupted by Microsoft 365. As a result, a number of Microsoft services from Azure via Exchange Online to Sharepoint Online were temporarily inaccessible to cloud users. However, for some customers using Microsoft's Azure SQL databases, this had far more serious consequences: they noted that their databases had been removed abruptly.
A user reported approximately: An Azure SQL database deleted and restored, but it was empty. Subsequently, affected Azure customers received a notification from Microsoft stating certain details about the cause. The Registry reportedly received such a letter from which the following quote would be taken:
"TDE Databases Must Be Removed"
"Automated Process Designed to Fire When Keys custom are removed from KeyVault. "
We are in the process of restoring a copy of these SQL databases from a recovery point in less than 5 minutes before the database is deleted.
Microsoft admits deletion
According to Microsoft, an automated process (probably during the Azure incident of January 29, 2019) resulted in Azure SQL databases were inadvertently deleted. Although Microsoft has again placed the databases on the same server and attempted recovery from a snapshot. However, one problem is that this recovery can take up to five minutes, the restored databases permanently missing transactions in a period of up to five minutes.
The problem was probably only for customers with Custom KeyVault keys for Use Transparent Data Encryption (TDE). TDE is recommended by Microsoft because it is designed to protect databases from attackers' malicious activity through encryption. Microsoft has asked the affected customers to open a support ticket in case of data loss affecting the processes or applications of the company.
Customers must intervene and be "compensated"
but if the recovered database is "more appropriate". This one can be renamed according to Microsoft by the customer in the original name and continue to be used. Microsoft wants to issue "Azure usage fees for all databases recovered for 2 months and for all databases for 3 months".