Deleting Amazon Glacier vaultsEdit

Created , updated

Amazon Glacier vaults are notoriously hard to delete, because they can’t be deleted until empty, getting a list of their contents is a multi-hour operation, and they can grow to be immensely large, requiring a lot of time to actually delete the contained archives.

For this purpose, I use the mt-aws-glacier tool.

Installing the tool

This is how I installed the tool on my OS X Mavericks box:

$ sudo cpan -i JSON::XS
$ sudo cpan -i Bundle::LWP # this is the p5-libwww-perl mentioned in the README
$ sudo cpan -i App::MtAws

Remember to recoil in horror as you realize that this pollutes /usr/local/bin (which Homebrew prefers to keep all to itself).

Prepare your credentials

$ cat > glacier.cfg <<HEREDOC

Prepare a list of vaults to be deleted

$ VAULTS="somevault

Request a list of archives in each vault

$ echo $VAULTS | while read VAULT; do
  echo "'$VAULT'"
  mtglacier retrieve-inventory --config glacier.cfg --vault $VAULT

Download the inventory

After 4 or 5 hours, you should be able to download the requested inventories:

$ echo $VAULTS | while read VAULT; do
  echo "'$VAULT'"
  mtglacier download-inventory --config glacier.cfg --vault $VAULT --new-journal $VAULT-journal.log

Note: If you have to retry this step at any point remember to delete your *-journal.log files; mtglacier will complain if you don’t.

Delete the archives from the vaults

$ echo $VAULTS | while read VAULT; do
  echo "'$VAULT'"
  mtglacier purge-vault --config glacier.cfg --vault $VAULT --journal $VAULT-journal.log

Delete the vaults

Supposedly you might have to wait 24-48 hours before doing this, but I suspect it may work immediately:

$ echo $VAULTS | while read VAULT; do
  echo "'$VAULT'"
  mtglacier delete-vault $VAULT --config glacier.cfg

Cleaning up after yourself

$ rm glacier.cfg *-journal.log

General observations

In my case I had about 700k archives and 8 vaults to delete, and it took about 3 days in all. I had to restart the operation two or three times, re-requesting the inventories (and waiting hours before they were ready) each time.