Delete all archives in an AWS Vault

aws_glacier_delete_vault.md

AWS Glacier: Delete vault

Follow these steps to remove all archives from an AWS vault. After this is finished, you will be able to delete the vault itself through the browser console.

Step 1 / Retrieve inventory

This will create a job that collects required information about the vault.

$ aws glacier initiate-job --job-parameters '{"Type": "inventory-retrieval"}' --account-id YOUR_ACCOUNT_ID --region YOUR_REGION --vault-name YOUR_VAULT_NAME 

This can take hours or even days, depending on the size of the vault. Use the following command to check if it is ready:

aws glacier list-jobs --account-id YOUR_ACCOUNT_ID --region YOUR_REGION --vault-name YOUR_VAULT_NAME 

Copy the JobId (including the quotes) for the next step.

Step 2 / Get the ArchivesIds

The following command will result in a file listing all archive IDs, required for step 3.

$ aws glacier get-job-output --account-id YOUR_ACCOUNT_ID --region YOUR_REGION --vault-name YOUR_VAULT_NAME --job-id YOUR_JOB_ID ./output.json

Step 3 / Delete archives

Set the following parameters through environment variables:

export AWS_ACCOUNT_ID=YOUR_ACCOUNT_ID
export AWS_REGION=YOUR_REGION
export AWS_VAULT_NAME=cvast-YOUR_VAULT_NAME

Create a file with the following content and run it:

#!/bin/bash

file='./output.json'

if [[ -z ${AWS_ACCOUNT_ID} ]] || [[ -z ${AWS_REGION} ]] || [[ -z ${AWS_VAULT_NAME} ]]; then
	echo "Please set the following environment variables: "
	echo "AWS_ACCOUNT_ID"
	echo "AWS_REGION"
	echo "AWS_VAULT_NAME"
	exit 1
fi

archive_ids=$(jq .ArchiveList[].ArchiveId < $file)

for archive_id in ${archive_ids}; do
    echo "Deleting Archive: ${archive_id}"
    aws glacier delete-archive --archive-id=${archive_id} --vault-name ${AWS_VAULT_NAME} --account-id ${AWS_ACCOUNT_ID} --region ${AWS_REGION}
done

echo "Finished deleting archives"

Acknowledgement

This tutorial is based on this one: https://gist.github.com/Remiii/507f500b5c4e801e4ddc

Thanks for sharing this, @veuncent

Here's a tweaked version of the script that processes in a stream (lower memory requirement for huge vaults), gives counts and timestamps, incorporates @joel1di1's fix, and uses AWS_PROFILE, if defined.

#!/usr/bin/env bash

file='./output.json'
id_file='./output-archive-ids.txt'

if [[ -z ${AWS_ACCOUNT_ID} ]] || [[ -z ${AWS_REGION} ]] || [[ -z ${AWS_VAULT_NAME} ]]; then
        echo "Please set the following environment variables: "
        echo "AWS_ACCOUNT_ID"
        echo "AWS_REGION"
        echo "AWS_VAULT_NAME"
        exit 1
fi

echo "Started at $(date)"

echo -n "Getting archive ids from $file..."
if [[ ! -f $id_file ]]; then
  cat $file | jq -r --stream ". | { (.[0][2]): .[1]} | select(.ArchiveId) | .ArchiveId" > $id_file 2> /dev/null
fi
total=$(wc -l $id_file | awk '{print $1}')
echo "got $total"

num=0
while read -r archive_id; do
  num=$((num+1))
  echo "Deleting archive $num/$total at $(date)"
  if [[ $AWS_PROFILE ]]; then
    aws --profile $AWS_PROFILE glacier delete-archive --archive-id=${archive_id} --vault-name ${AWS_VAULT_NAME} --account-id ${AWS_ACCOUNT_ID} --region ${AWS_REGION}
  else
    aws glacier delete-archive --archive-id=${archive_id} --vault-name ${AWS_VAULT_NAME} --account-id ${AWS_ACCOUNT_ID} --region ${AWS_REGION}
  fi
done < "$id_file"

echo "Finished at $(date)"
echo "Deleted archive ids are in $id_file"

I'd recommend naming it delete-archives.sh and running it in the background on a machine that's going to be on a network for a long time, e.g.:

chmod 755 delete-archives.sh
nohup ./delete-archives.sh > delete-archives.log 2>&1 &
tail -f delete-archives.log

Nice, thanks for sharing @cwilper !

if [[ -z ${AWS_ACCOUNT_ID} ]] || [[ -z ${AWS_ACCOUNT_ID} ]] || [[ -z ${AWS_ACCOUNT_ID} ]]; then

Should be :
if [[ -z ${AWS_ACCOUNT_ID} ]] || [[ -z ${AWS_REGION} ]] || [[ -z ${AWS_VAULT_NAME} ]]; then

Thanks @joel1di1 ! I updated the script. (sorry for the late reply, I didn't get a notification when you posted this)

Hi @cwilper
I'm not returning anything from the output.json file. There are archive-ids in there but it returns "Getting archive ids from ./output.json...got 0". Any thoughts

Thank you @cwilper, your script let us to delete a 128TB vault (24GB json in file size).

Update:
@@ my 1G Centos VM estimated 28 months to delete all the files, (1.5 sec/file deletion)
It will be great if anyone has a multiple thread enabled script :-)

Finally I have combined multi thread idea with @cwilper script, hope anyone can use it. The best of the script is that it can be utilise all the machine RAM (as low as 2GB) and the Max CPU thread to carry out the task without crash the machine while doing it.

#!/usr/bin/env bash

file='./output.json'
id_file='./output-archive-ids.txt'

if [[ -z ${AWS_ACCOUNT_ID} ]] || [[ -z ${AWS_REGION} ]] || [[ -z ${AWS_VAULT_NAME} ]]; then
        echo "Please set the following environment variables: "
        echo "AWS_ACCOUNT_ID"
        echo "AWS_REGION"
        echo "AWS_VAULT_NAME"
        exit 1
fi

echo "Started at $(date)"

echo -n "Getting archive ids from $file..."
if [[ ! -f $id_file ]]; then
  cat $file | jq -r --stream ". | { (.[0][2]): .[1]} | select(.ArchiveId) | .ArchiveId" > $id_file 2> /dev/null
fi
total=$(wc -l $id_file | awk '{print $1}')
echo "got $total"

num=0
while read -r archive_id; do
  num=$((num+1))
  echo "Deleting archive $num/$total at $(date)"
  aws glacier delete-archive --archive-id=${archive_id} --vault-name ${AWS_VAULT_NAME} --account-id ${AWS_ACCOUNT_ID} --region ${AWS_REGION} &
  [ $( jobs | wc -l ) -ge $( nproc ) ] && wait
done < "$id_file"

wait
echo "Finished at $(date)"
echo "Deleted archive ids are in $id_file"

You can freely to change the variable "$( nproc )" for the CPU thread will be used. We tested this above script will already getting the max machine thread number and give the best performance a already.
Just execute it as @cwilper said.

chmod 755 delete-archives.sh
nohup ./delete-archives.sh > delete-archives.log 2>&1 &
tail -f delete-archives.log

Nice @goodspeedal, looks like a good approach to get better throughput.

Great job, thanks @cwilper @goodspeedal !

This was very helpful. Anyone else feel like a Glacier vault is the digital equivalent of Hotel California...

They gathered for the feast
They stab it with their steely knives
But they just can't kill the beast

:D

This was very helpful. Anyone else feel like a Glacier vault is the digital equivalent of Hotel California...

They gathered for the feast
They stab it with their steely knives
But they just can't kill the beast

Yes! I'm kind of sorry I ever stored anything on Glacier.

Very helpful! Thank you for posting this!

awesome guidance here!

Hi @veuncent. Thanks for this. I've borrowed from it and created what I think it's a more end-to-end version of it. You just need to pass account_id, aws_region and vault_name and it will do the rest. Hope it's useful to yourself and others.

https://gist.github.com/dmabamboo/e4c0c4a356116c026a86628c11864ab0

Works like a charm, thank you! Btw the need to have this script is one of the reasons I'm moving to Azure Archive Storage; the other is the half of the price. :)

Oh, great. I just wrote this https://github.com/guillem/glacier-delete and now I find out you people already have made the two biggest improvements my version is missing (multithreading and "end to end" process) xD Anyway, here it is.

Many thanks to everyone here!!

This is awesome. Thanks!!! This saved me from a wasted morning writing code I did not want to write just to exit an ecosystem I no longer need or want.

Similar to @m-leishman I'm getting the same result using the combined code posted by @goodspeedal

Getting archive ids from ./output.json...got 0

echo -n "Getting archive ids from $file..."

• echo -n 'Getting archive ids from ./output.json...'
  Getting archive ids from ./output.json...if [[ ! -f $id_file ]]; then
  cat $file | jq -r --stream ". | { (.[0][2]): .[1]} | select(.ArchiveId) | .ArchiveId" > $id_file 2> /dev/null
  fi
• [[ ! -f ./output-archive-ids.txt ]]
  total=$(wc -l $id_file | awk '{print $1}')
  ++ awk '{print $1}'
  ++ wc -l ./output-archive-ids.txt
• total=0
  echo "got $total"
• echo 'got 0'
  got 0

num=0

• num=0
  while read -r archive_id;

Me too on the 'got 0' output. Weird.

Edit - You've got to install jq and in my case (Mac) had to rename the downloaded binary to jq and add it to the path. https://stedolan.github.io/jq/download/

I tried a few of the linked items above, but came back to this one as it's the simplest. Even so, I still don't know if it's working - I suspect Glacier has to update the inventory again before I know, which I guess I need to trigger. Will report back for others like me whom are not developers, trying to figure out this nightmare. Never again Glacier, never again!

Reporting Back
Yeah, so once deleted you have to wait overnight and then check the next day for the deleted files to show up in the GUI - for me, one vault was able to be deleted straight away and the other had only a few files left. So I'm in the midst of running the process again for the vault with some files remaining, but it looks like that'll do it! 6-7 years of not being able to get rid of these files lol. They make it so easy to put them in!

Thank you all, I am now on my way to deleting my ghost vaults. Yes, it does feel like Hotel California unfortunately ...

Thank you so much for this, Synology backup created over 200K archives when I backed up 8TB of data, was contemplating my next week of copy pasting archive IDs until I found your script!

I am still getting Getting archive ids from ./output.json...got 0 with this in Ubuntu. I have tried several of the latter scripts.

I have jq installed, configured, in path and can run simple commands that use jq, so I know that is working. I have regenerated the output.json as well

The scripts also work fine in AWS CloudShell, but with a 120mb output.json file the timeout there (20mins) negates this as an option.

Desperate to get these 270k archives gone.

(Edit - the first script is working, slowly, but working)

If the completed ids file './output-archive-ids.txt' (or whatever you've set id_file to) already exists, it won't read the input file and you'll see "got 0":

You have to delete or rename that file each time you run it.

Thanks much!

Thanks for this! I changed $( nproc ) into 1 on my Macbook, otherwise the script would consume too much CPU (doesn't recognise nproc).

Also, I had a few network hiccups - the script missed some - so I created another job for the remaining archives to delete. These id's get appended to id_file by default. So either (re-)create this file in the script or remove output-archive-ids.txt before the next run :)

Thanks for sharing this, super helpful. I was able to get $( nproc ) working on MacOS by installing coreutils (e.g. brew install coreutils). May be helpful for others running a large archive deletion on a Macbook.

Good script, thanks a lot!

Thank you!

Here is script that using xargs with 8 parallel process:

#!/bin/bash

file='./output.json'

if [[ -z ${AWS_ACCOUNT_ID} ]] || [[ -z ${AWS_REGION} ]] || [[ -z ${AWS_VAULT_NAME} ]]; then
        echo "Please set the following environment variables: "
        echo "AWS_ACCOUNT_ID"
        echo "AWS_REGION"
        echo "AWS_VAULT_NAME"
        exit 1
fi

jq -r .ArchiveList[].ArchiveId < $file | xargs -P8 -n1 bash -c "echo \"Deleting: \$1\"; aws glacier delete-archive --archive-id=\$1 --vault-name ${AWS_VAULT_NAME} --account-id ${AWS_ACCOUNT_ID} --region ${AWS_REGION}" {}