Glacier is only 1 cent per gigabyte per month, so it means you can store the entire contents of a 250 GB SSD for about $2.50 per month.
- You can only restore 5% of your total storage tally (ie. 12.5 GB in the case of the example above) each day, unless you want to pay extra, and restoring is slow. In other words, this is a place to sling data for a rainy day, where you’re only likely to need it in the event that disaster truly strikes.
- The open source arq_restore tool hasn’t been updated to support back-ups made to Glacier yet; there is a ticket for that on the arq_restore GitHub page.
- There is no convenient migration path for existing S3-based backups; you effectively have to delete them and then back-up all over again to Glacier.
Currently I’ve allocated a $30/month budget for my S3 backups, and it’s required some micro-managing of excluded files to make sure that the stuff I want gets backed up and irrelevant cruft (temporary files, log files, cache files etc) doesn’t get transferred. In all the time I’ve been using Arq (around a year now, I think), I’ve needed to use its restore functionality on one occasion to save my hide.
So for me, switching to Glacier seems like a no-brainer. I am hopeful that the arq_restore tool will get an update to enable restoration in a world without Haystack Software. In the meantime, upgrading and switching to Glacier means taking a gamble on that front and risking permanent vendor lock-in.
I’m also hopeful that some of the sharp edges around Glacier support will be smoothed over with future updates.
Although it will be too late for me by the time it happens, I’m hopeful that we’ll see nice migration support. I also see somewhat of a gaping hole in the "budget enforcement" department on the Glacier side, where the cheapness of it basically means that there is no budget enforcement yet. That might be ok for now, but as the snapshots start piling up and the costs start to mount I can see it becoming important.