glacier-utils

a few tools I threw together to be able to archive directories to Amazon Glacier

tree-hash

a bash implementation of their tree-hash algorithm.

Usage:

tree-hash [-b] filename

  -b  print the output hash in binary form

tree-hash contains 2 "subcommands" that can be run individually:

tree-hash-reduce

Implements the 'merge level' portion of the tree hash algorithm. Given a starting directory of leaf nodes, will combine 2 hashes at a time, and repeat this process until only 1 hash is left over.

Usage:

tree-hash-reduce directory

A common task is building the root treehash from a list of subnode hashes. This is implemented in the tree-hash command as:

mkdir -p hashes
cp hash.* hashes
tree-hash-reduce hashes
xxd -ps -c 100 "hashes/$(/bin/ls -1 hashes | head -n 1)"

tree-hash-combine

Almost never need to run this directly.

Takes 2 binary hash files (left-file and right-file), concatenates the values together and sha256 hashes the result. The resulting hash is written back to left-file and right-file is deleted.

Usage:

tree-hash-combine left-file right-file

glacier-archive

Archives a directory to Amazon Glacier.

Usage:

glacier-archive dirname

glacier-archive-prepare

Prepares a directory for archiving:

1. creates an archive.$dirname sidecar directory that holds the prepared data
2. tar the directory and split the result into equally sized chunks
3. run tree-hash on each chunk
4. run tree-hash-reduce on the resulting hashes to get the tree-hash of the whole archive

Usage:

glacier-archive-prepare dirname

glacier-archive-upload

Uploads a directory (after running Glacier-archive-prepare to Glacier using the multi-part uploader. After uploading, autotags the original (unprepared) directory with the metadata returned from the Glacier API.

Usage:

glacier-archive-upload dirname

Note: glacier-archive-upload should take the same dirname argument as glacier-archive-prepare and not the prepared archive.* directory.