Moving from Glacier to S3 Glacier Deep Archive

Question

I have a lot of files located in Amazon Glacier (the pre-S3 version, and have been uploading files using FastGlacier for years.) I would now like to move these files to the S3 Glacier Deep Archive storage class to take advantage of better pricing, and am trying to figure out the best way to do so. (It is my understanding that the pre-S3 Glacier does not offer the deep archive storage class, but I am happy to be wrong.)

Is there any way I can restore my files from Glacier directly to an S3 account/bucket/whatever, so I can avoid the bandwidth usage associated with downloading everything to my house, only to re-upload it to the cloud? Or is that my only option?

Answer 1

No.

The original Amazon Glacier service was great for its low cost, but was difficult to use. Most requests required you to come back minutes (or hours!) later to retrieve the result. It was almost impossible to use without a tool, such as the one you mentioned.

Then the Amazon S3 team introduced the ability to specify Glacier as a storage class in S3, and they would take care of the difficult bits. This gave the low cost a better interface.

More recently, the Glacier Deep Archive storage class brought even lower costs than Glacier itself, so there's hardly any reason to use Glacier directly any more. Unless, of course, your files are still in there, as is your case.

Unfortunately, there is no mechanism to move from "old Glaicer" to "S3 Glacier". You would need to extract the archives and then upload them to S3 (either specifying Glacier Deep Archive as the storage class or use S3 Lifecycle rules to change storage class). You would need to do this yourself, preferably from Amazon EC2 to make things faster and avoid Data Transfer charges. Perhaps you put your FastGlacier tool on a Windows EC2 instance and do it from there?

Answer 2

You can fire up an EC2 or Lightsail instance, with a public IP address, in the same AWS region as your Glacier vault and new bucket and do all the download/upload from there, avoiding the bandwidth costs since the charges are for the traffic to leave the AWS region -- and with a compute instance in the same region, that wouldn't apply.

There is no mechanism for directly transferring content from one service to the other, and you are correct... the legacy Glacier service does not appear to support deep archive.