Hacker News Reader

Comment by buildzr | Hacker News Reader

> when you loose your data you have only 30 days to get it back before Backblaze deletes it as well!

Uhhh... this one could be a significant issue for me right now. I've got a laptop that's been offline for about 2 months now due to a motherboard failure due to liquid damage. I haven't bothered dumping the drive yet because I figured even if it's got issues Backblaze has a copy.

You're telling me that if I login to my account right now, the data is gone? If so, they really need to make that one more obvious.

EDIT: Just checked Backblaze, still looks like the data is there after 70+ days. Has this policy changed at some point?

View on HN

Disclaimer: I work at Backblaze.

> Has this policy changed at some point?

It is more complicated than just "after 30 days" (and always has been). If your laptop is entirely offline (or you simply uninstall the client from your laptop), then the policy is ACTUALLY that we keep your backup for 6 months as long as you keep paying your bill. But to be honest, it's more like a year or two. The 6 month policy is we guarantee the backup will be preserved, not that we will immediately go out and delete it at exactly 6 months and 1 day.

This is a completely different situation than files you delete but the backup continues. For cost reasons, Backblaze purges the files that you deleted from your local drive after 30 days. HOWEVER, with this new 7.0 release you can pay a little more and increase that retention time to one year, or forever. This was a highly requested feature for the situation you ran into.

Or, if you have a hard drive that you disconnect and aren't willing to reconnect for more than 30 days. Since your backup is continuing, Backblaze assumes the drive will never come back, so the files are then deleted to save money in the datacenter. Of course, this changes with the 1 year retention policy, you can unplug the drive for up to 1 year and still dial back time and restore all your files.

> HOWEVER, with this new 7.0 release you can pay a little more and increase that retention time to one year, or forever. This was a highly requested feature for the situation you ran into.

I'm glad to hear that Backblaze has finally made such an option available. However:

> For cost reasons, Backblaze purges the files that you deleted from your local drive after 30 days.

As someone who once almost lost his PGP key and had to recover it from old physical backup media, I'd like to point out that that is not a "backup" service. It's sort of like a lazily expiring mirror, but it's definitely not a backup service.

Do you look at all of your files every 30 days? How long would it take you to notice that a random file had disappeared from the filesystem 7 layers deep? Have you ever needed to restore a years-old file?

(You need not explain why you do it; I understand about users who could use it as a cloud storage service by deleting files after they're backed up. The point remains that it's not a backup.)

What's your criteria for backup? I completely agree that something like RAID is not backup, but 30-60 days seems comparable to other backup services and covers most scenarios. The old-school manual process of swapping a USB drive at work every week/month has the similar retention.

The only consumer-facing system I've used with multi-year retention is Time Machine. Every few years it has trouble "verifying" the backup and I have to start over (it also deletes old backups when you run out of space). Right now my backup only goes to August.

Backup means that if it takes me 6 months to realise that the 2014 finances folder had been accidentally deleted, I can still recover it.

I'd love to see BackBlaze offer the option to select "mission critical" folders that have super-long-life (Time Machine-like) version retention. Give everyone a few gigabytes of this for free, charge a premium to increase that space.

Though that sounds tricky to set up all the UI for. For now you can set up a separate backup of that stuff. Set B2 as the backend and you get 10GB free.

Maybe—and the UI could be little more than a script that does exactly what you described.

The point of making it a UI is, first and foremost, to get people thinking about this question. And it gives people the option to acheive what you just described without stumbling upon a post like yours on Hacker News.

A post like mine? But it's your post that has the idea...? The only thing I said about how to make it possible is "separate backup", which anyone can think of in two seconds and is only a hint of a tenth of an explanation. I don't understand.

> What's your criteria for backup?

Something like Attic, Borg, CrashPlan, Restic, etc, that allow snapshot retention periods, like "1 per year for the last X years, 1 per month for the last X months, one per day for the last X days".

> I completely agree that something like RAID is not backup, but 30-60 days seems comparable to other backup services and covers most scenarios.

That doesn't protect against bitrot. e.g.

1. File is backed up (snapshot A).

2. File is slightly corrupted by bitrot, cosmic ray, etc.

3. File is modified by user, corruption is unnoticed.

4. File is backed up again (snapshot B).

5. "Backup" service deletes snapshot A.

6. User discovers corruption.

7. User looks to restore earlier versions until an uncorrupted one is found.

8. User discovers that all available snapshots were made after corruption happened.

Or replace "file" with "directory" and "bitrot" with "accidental file deletion" and the user still suffers from data loss.

Mirrors are not backups, and a few revolving snapshot slots is effectively a mirror.

What you're referring to, I'd call a historical archive, containing many backups.

When I think about a backup, I'm thinking in terms of recovering last good state after a drive failure or catastrophic filesystem corruption. I don't tend to think of a backup as implying a deep history unless that part is explicitly stated. That distinction was easier to notice back in the days of backing up to tape or optical discs - you don't expect each tape/disc to contain a version history, just a single snapshot, and you don't expect your collection to retain long-gone files unless it's an ever-growing pile of tapes/discs.

In that mindset, it's not reasonable to expect that a backup service necessarily provides the full historical archive.

> you don't expect your collection to retain long-gone files unless it's an ever-growing pile of tapes/discs.

Rule number whatever of backups: Don't discard old backups. Bitrot occurs, and people make mistakes, and newer backups can have flaws that older ones don't, which will go undiscovered until it's too late.

We're not talking about piles of tapes, we're talking about virtually unlimited disk space. State-of-the-art backup software chunks and deduplicates data and stores snapshots of directory trees. For most cases, there's no reason not to keep a subset of old snapshots, and many reasons to keep them.

I should have explained the 30 days better. I would edit my comment if I still could.

But yeah, since I kept the computer online and Backblaze app running _after_ I lost the files, Backblaze marked them up for deletion in those 30 days. To be fair, I didn't try to contact them and ask to help me, maybe they would.

I spent several days trying to download the backup and then I gave up and ordered the drive.

Good to see that you have a new retention policy since then.

Reply on news.ycombinator.com