KV Sites on Facebook

Backups

Kae Verens ~ Saturday 31st March, 2012

Today is World Backup day, so I thought I'd talk a bit about how we do our backups.

KV Sites' services are spread out currently over 4 separate hosting solutions. This means that if there was ever a problem with one of them, there are the other three there to continue the service. This has never happened to me in ten years of this game, and I don't ever expect it to, but it is never wise to put all your eggs in one basket.

Every day, an automatic script logs into each server using a program called rsync, and synchronises a local copy of that server.

This means that even if one of the online services goes completely belly-up, there is always a copy of the server ready to upload in another location.

The backups are stored on a large ZFS mirrored array.

ZFS is a file system which lets you use a load of hard-drives together to form one large file partition. We have ours set up with a number of 1TB and 2TB hard-drives to form one single large file system, which we use for backups.

We also mirror the hard-drives. This means that for every 1TB in the system, there is another one which is an exact clone of that first one. If one of the hard-drives ever breaks down, the other one carries on and no data is lost.

ZFS also automatically creates a checksum of every file that's added to the system, and checks it every time there is anything done to the file. A checksum is a long number which is generated by performing a mathematical "hash" algorithm on the file. It's a very good way of checking that the file is exactly the same way now as it was yesterday. This means that if something unlikely happens such as a file being changed by some of its memory being hit by cosmic radiation, the file system will notice, will check the mirror, and will correct itself automatically.

So in summary, your data is very safe with KV Sites because:

  • We use a number of separate datacentres to host our servers
  • We backup everything daily
  • Our backup hard-drives are mirrored
  • Backup files are checked for corruption automatically and corrected if anything is found