1 Answers
"Snapshots" are not truly analogous to Full and Incremental backups, in that earlier snapshots are not required to be available to do a full "restore" from any snapshot. All snapshots are a "full" recovery point. The "snapshot" is actually a catalog of all of the blocks of data written to S3 during the snapshot process. As long as a block of data does not change on the EBS volume, it wont need to be stored in S3 again, but it will still be cataloged in each snapshot. Likewise, as long as a snapshot exists that tracks a particular block of data, that block of data will remain in S3.
So, if you have taken a series of snapshots, but the data on the EBS volume hasn’t really changed, then deleting that first snapshot might not cause any data to be removed from S3 at all, as later snapshots might still reference blocks that had been stored during the first snapshot operation.
If a block of data has been deleted or modified on the EBS volume, then as long as a snapshot remains from a time when that block had not yet been modified, that older data will persist in S3. When the final snapshot referencing a particular block of data in S3 is deleted, only then will that block of data be deleted from S3.
Hope that clears things up.
Thank you! That’s really helpful. To check I have fully understood, let me ask one further clarification question.
There is a distinction between a snapshot and the blocks of data that are the copy of the EBS volume. The snapshots are signposts to the blocks, with re-use among snapshots where no changes to blocks of data have occured?
Yes…the blocks in S3 are re-referenced by later snapshots – presuming said blocks are still unchanged on the EBS volume.