How does AWS transfer S3 objects to Glacier archives when you use lifecycle archive rules?

Question

Amazon Web Services (AWS) S3 allows you to automatically transfer/archive objects from S3 to Glacier.

Nowhere that I can find does anybody explain how that transfer happens.

What Glacier vault do S3 objects archive to? Does the lifecycle policy set any sort of description on the Glacier archives it creates? Does it create an archive per S3 object?

Answer 1

Q: Can I use Amazon Glacier APIs to access objects that I've archived to Amazon Glacier?

Because Amazon S3 maintains the mapping between your user-defined object name and Amazon Glacier's system-defined identifier, Amazon S3 objects that are stored using the Amazon Glacier option are only accessible through the Amazon S3 APIs or the Amazon S3 Management Console.

https://aws.amazon.com/s3/faqs/#glacier

To put this in slightly simpler terms, S3 doesn't create Glacier archives that you own or can manipulate. S3 creates Glacier archives that S3 owns and manages.

Your only interface to these objects is through S3, which makes requests to Glacier on your behalf. So, for your questions, the answer for each one is essentially the same:

It doesn't matter. The archives are managed by S3 and are not user-accessible via the Glacier API or console.

In fact, it appears that S3 isn't accessing "your" Glacier at all. No permission grants are needed to allow S3 to access "your" Glacier (by contrast, when other services create objects in "your" S3, you have to explicitly grant permission). Glacier also doesn't appear to know or care that the objects from S3 are from your AWS account. On the monthly billing, S3 charges you for the Glacier storage -- not Glacier.

It is modeled as if S3 has its own AWS account, and S3 is a Glacier customer. S3 creates these objects in Glacier, then passes the storage charges back to you.

That's an oversimplification, I'm sure, but your ability to interact with Glacier, directly, when S3 is archiving objects there for you is non-existent... thus explaining why you can't find answers to these questions: because the answers are not known, not public information, and not relevant or important.

S3 objects archived to Glacier are still S3 objects . You can still send a HEAD HTTP request to S3 and see the object's metadata. You just can't download the object until you initiate a restore from Glacier, which S3 manages, and the restore operation completes. When the object is restored, it now persists in Glacier and has a temporary, downloadable copy held in S3 REDUCED_REDUNDANCY storage, which S3 will discard again after the number of days you specified when you asked S3 to restore the object.