admin:backup:start

This is an old revision of the document!


Backup

Statistically speaking, you will eventually lose something that can't be replaced. Family photos, your thesis, the notes from that meeting, the song you recorded with your dad. It's just a matter of when, what, and how much you lose due to

  • a drive failure,
  • fire,
  • flood,
  • ransomware,
  • or theft.

What should I back up?

You should have backups of any data that A) is useful to you and B) can't be reliably replaced (rescanned, re-downloaded, or get a copy from uncle JimmyJoBob).

RAIDs are "fault tolerant" in that they can survive some drive problems or failures, but fault tolerance is not considered having an additional copy of the data and doesn't count as a backup as your system is still prone to most of the scenarios listed above.

The first rule is 3-2-1. Tried and true. You should have at least

  • Three copies of your data,
  • on two different media,
  • at least one of which is stored offsite from your computer.

Having two copies of the data on the same hard drive (same media) is NOT a backup. If the single drive fails, you've lost both copies. Likewise, having two copies in the same email account (same media) is NOT a backup. If that email service craps out, you've lost both copies.

You don't want a single failure to wipe out both copies. They should be on a cloud storage, removable media, NAS, etc that is not a permanent part of your computer.

If you have 50 copies of your critical files, stored on 50 different media, but they are all at your home/office… well what if the house burns down or floods? What if someone steals it all? These things can happen. So at least ONE of those copies should be offsite. This could include:

  • remote cloud storage,
  • a remote server,
  • or even a hard drive that you store somewhere far from your computer (home, office, bank lock box, etc).

Example

Copy 1: Local HDD. Duh.

Copy 2: Remote storage. For this, I recommend using a FOSS option like Nextcloud, duplicity, rdiff-backup, borgbackup, rsync coupled with a remote server you own / have access to, or a proprietary program such as Crashplan, SpiderOak, Carbonite. These can all be configured to do incremental backups at set times, near-real-time backups, and versioning.

A real-time sync is NOT a remote backup if it doesn't have versioning. That is, if deleting the local file also deletes the remote file in real-time, and there is not way to get an older archived version of the file, that is not a backup.

Copy 3: For this, since we've already satsified the off-site aspect, we can just do a second copy on a local NAS, removable media (HDD/DVD/USB). This allows fast recovery if something as mundane as a local file corruption or primary HDD crash happens.

Okay, NOW you have backups. You know what else is not a backup? One that hasn't been tested. If you haven't tested your ability to recover data from a backup, you have no verified backup.

How do you test this? Easy enough. Back up some trivial data you don't care about to all three locations from time to time. Then delete it from your primary drive.

Now, and this is critical, wait a reasonable amount of time. Think about how long it might be before you realize that a file you needed has been corrupted or is missing. A few hours? A few days? You want your test to mimic that scenario as much as reasonably possible.

Now try to recover a copy from each of your backups. If you can do so, you are golden. If not, then one or more of your backup locations is not reliable and should be replaced or fixed.

  • User data
    • Windows: C:\Users
    • Linux: /home
  • MBR / GPT
  • partition layouts
  • configuration files
  • Last modified: 2020-04-02 10:58