Using s3 Glacier for backups

Now that Duplicati is Beta, I’m planning to use it for my offsite backups to s3. As these backups are not going to be accessed unless I lose all of my local discs, I think it makes sense to use Glacier.

After going through Duplicati’s guide to use Glacier, I have the following questions:

  1. Assuming Amazon does not corrupt or lose any of my files, what, if any, are the risks of running with --no-backend-verification?
  2. What will I have to do if I wish to make a full restore of my data?

Glacier is a bit problematic, due to the restore process and it has not been fully automated (nor integrated in Duplicati). It works by using the S3 life-cycle rules to move files to Glacier. This means that Duplicati can no longer verify that all is working as expected, so you need to “trust” it, which is a bad idea.

To test the restore, you basically have to recall the files from Glacier. If you can get them moved back into S3, you can test the restore.

I generally recommend not using Glacier with Duplicati for these reasons.

Also, there is a longer discussion on the topic here: Amazon Glacier support · Issue #701 · duplicati/duplicati · GitHub

1 Like

Thanks for the quick reply! Would it then be o.k. to use Standard Infrequent Access or should I just stick to Standard?

If you use IA there is a fee if you access the files too quickly. Duplicati will do a test by downloading a file after it has been uploaded and then checking if it is stored correctly.

This will incur a penalty fee.

You can use the advanced option --backup-test-samples=0 to skip the download, but still check that the list of remote files is correct.

1 Like

But with a Lifecycle Policy that moves dblocks to IA after 30 days, that should not be a problem, right?

I do not know the AWS pricing well enough to help with that :frowning: