2

I'm planning to back up multiple websites to Glacier at midnight each day.

My current system is to create a tar.gz of the whole of /var/www (i.e. all the sites), and to copy that to a different server.

I'm assuming I would use these backups if the whole server failed, though perhaps I'd occasionally want to retrieve individual files.

Should I continue in the same way (tar.gz of all per day in one Vault), or have different tar.gz per site in the same vault, or different tar.gz per site each in a vault per site?

Rich Bradshaw
  • 177
  • 1
  • 8
  • 1
    Glacier is nice, but really isn't best for backups since they charge quite a bit to access data (it's more for archiving data...like for HIPPA compliance and so forth). S3 might work, too. – Nathan C Dec 08 '13 at 15:21
  • 1
    I might be wrong, but the backups I'm using are around 1GB per day – I also keep local versions. If we had a complete failure I'd be pulling say, 2GB. That means I'd get 1GB free, then the next GB would cost $0.12. If it were all one tar.gz, I'd need one request call. Is that correct? Seems really cheap when using around 1-10GB per backup? – Rich Bradshaw Dec 08 '13 at 16:02
  • 1
    For small backups you'd be fine. For terabyte-size backups glacier is not cost-effective compared to S3, but for smallish backups it's cheap. – Nathan C Dec 08 '13 at 18:00

1 Answers1

2

The "best of both worlds" approach would be the following:

  1. Create a S3 bucket for backups
  2. Configure a lifecycle rule, so items older than 2 weeks (for example) are automatically transitioned to Glacier class storage.
  3. Plan your backups as simple file syncing tasks using S3 for destination. You have plenty of tools to automate this.

By doing it this way, you will have a decent "instantly ready" backup in S3 for the last weeks, with much more advanced copy tools (sorry, Glacier) and you will benefit from the lower costs of Glacier and the S3 interface to retrieve items, much friendlier than Glacier's.

ma.tome
  • 1,179
  • 8
  • 15