Test restore from object storage, putting performance tier in maintenance mode.

[ConradGoodman]

I did a test restore of a DB from AWS S3 object storage.

Having searched on here, I used the method of putting the performance tier into maintenance mode in order to force an object storage restore.

I restored application items (SQL) from a VM.

The VM disk was 230GB, yet Veeam seemed only to download a v small amount of data from AWS?

I published a single SQL DB to the target SQL server.

Can someone explain exactly what happens here? The backup file is mounted over iSCSI from the Veeam server to the destination SQL server.

Does this mean the backup file in object storage is mounted via the Veeam server, without actually downloading it? If so this is excellent, but I must confirm.

Seeing as the disk was 230GB, there is no way that was all downloaded, yet the whole disk is mounted in my target VM.

[veremin]

Does this mean the backup file in object storage is mounted via the Veeam server, without actually downloading it? If so this is excellent, but I must confirm.

Confirmed. Thanks!

[Gostev]

Seeing as the disk was 230GB, there is no way that was all downloaded, yet the whole disk is mounted in my target VM.

Think of it as a stub of the disk: a virtual device with no "physical" content behind. The actual content of disk blocks are retrieved from object storage in real-time as they are requested (as SQL Server reads them from a database file, for example when you run a SQL query). While from the OS perspective, it is just reading the data requested by an application from a block device normally, except the I/O latency is a bit higher than it would be for a local disk.

[ConradGoodman]

Thanks for the information guys, very helpful