Amazon S3 Glacier

Amazon S3 Glacier (glacier) on fakecloud: a complete 33-operation implementation (100% conformance) with a real data plane — vaults, archive upload/delete with SHA-256 tree hashes, multipart uploads, retrieval and inventory jobs, vault lock, notifications, access policy, tags, data-retrieval policy, and provisioned capacity. restJson1.

fakecloud implements Amazon S3 Glacier as a restJson1 service. All 33 operations ship with 100% conformance against AWS's own Smithy model, backed by account-partitioned state that persists across restarts in persistent mode.

Unlike a control-plane-only mock, Glacier on fakecloud has a real data plane: uploaded archive bytes are stored, a real SHA-256 tree hash is computed, and retrieval jobs return the exact bytes you uploaded.

Account-scoped paths

Glacier's REST API scopes every request under an account id (/{accountId}/vaults/{vaultName}/...). fakecloud accepts the literal - that AWS documents as "the caller's account" — and, permissively, any account id — mapping it to the authenticated caller's account. Custom request headers (x-amz-archive-description, x-amz-sha256-tree-hash, x-amz-part-size, Content-Range, Range) are honoured, and ids are returned in response headers (x-amz-archive-id, Location, x-amz-multipart-upload-id, x-amz-job-id, x-amz-lock-id, x-amz-capacity-id) exactly as the model's @httpHeader bindings declare.

Archives round-trip end to end

UploadArchive stores the real request bytes and computes the archive's SHA-256 tree hash — Glacier's binary hash tree over 1 MiB chunks, which collapses to a plain SHA-256 for small archives. When the caller supplies an x-amz-sha256-tree-hash header it is validated against the computed hash; a mismatch returns InvalidParameterValueException.

A retrieval then returns the same bytes:

  1. InitiateJob with {"Type": "archive-retrieval", "ArchiveId": "..."} creates a job (HTTP 202) and returns its id in x-amz-job-id.
  2. The job settles to Succeeded on the first read (DescribeJob, ListJobs, or GetJobOutput), mirroring the eventual consistency of the real multi-hour retrieval — without the wait.
  3. GetJobOutput streams the stored archive bytes back, honouring a Range header for partial retrieval (HTTP 206 with a Content-Range).

DeleteArchive removes the bytes; the vault's NumberOfArchives and SizeInBytes update accordingly.

Inventory jobs

InitiateJob with {"Type": "inventory-retrieval"} builds a JSON inventory of the vault's archives (id, description, creation date, size, tree hash); GetJobOutput returns it as application/json. ListJobs supports the statuscode and completed filters plus marker/limit pagination.

Multipart uploads

InitiateMultipartUpload (with an x-amz-part-size that must be 1 MiB times a power of two) starts an upload; UploadMultipartPart stores each part keyed by its Content-Range; ListParts and ListMultipartUploads page through the in-flight state; and CompleteMultipartUpload assembles the parts in order into a stored archive, validating the declared x-amz-archive-size and the combined tree hash. AbortMultipartUpload discards the upload.

Vault lock

The vault-lock state machine is real: InitiateVaultLock moves a vault from unlocked to InProgress and returns a lock id with a 24-hour expiry; CompleteVaultLock with that id transitions it to Locked; GetVaultLock reports the current State, Policy, CreationDate, and ExpirationDate; and AbortVaultLock cancels an in-progress lock (a completed Locked vault cannot be aborted). A stale in-progress lock expires automatically after 24h.

What else is covered

  • Vault notifications — SetVaultNotifications / GetVaultNotifications / DeleteVaultNotifications persist the SNS topic and event list.
  • Vault access policy — SetVaultAccessPolicy / GetVaultAccessPolicy / DeleteVaultAccessPolicy round-trip the policy document.
  • Tags — AddTagsToVault / RemoveTagsFromVault / ListTagsForVault (the tags POST is discriminated by the operation=add|remove query param).
  • Data-retrieval policy — GetDataRetrievalPolicy reports the account policy (defaulting to a BytesPerHour rule), SetDataRetrievalPolicy persists a custom one.
  • Provisioned capacity — PurchaseProvisionedCapacity returns a capacity id; ListProvisionedCapacity lists the account's units.

Errors

Glacier returns its own {"code", "message", "type"} error envelope with the model's declared HTTP status: ResourceNotFoundException (404) for a missing vault, archive, upload, or job; InvalidParameterValueException (400) for a malformed parameter or tree-hash mismatch; and MissingParameterValueException (400) for an omitted required parameter.