Home / server / Questions / 349474
Accepted
MTAdmin
Asked: 2012-01-13 15:47:00 +0800 CST

Tape Backup Scheme - Trying to save tapes

  • 772

I have inherited a tape backup scheme that currently doesn't wash any tapes. Expensive, but fine. We currently run Fulls every quarter, which become our offsite. We then run incrementals in the interim. The fulls are permanently archived offsite.

I would like to switch to a more traditional scheme to save tapes.

However, I can't account for recovering data indefinitely that is deleted by a user (or becomes corrupt) between the full backups. e.g. I have a large ESRI data set of 400 files. A full backup is made on Jan 1. User processes the dataset for several weeks then at some point archives it to network storage. Unbeknownst to the user, one of the files in the dataset becomes corrupted, rendering the entire dataset unusable.

Is there any washable backup scheme that could recover from this type of lost file indefinitely, or is my issue that I need to define a time frame within which I can provide recovery?

Just curious if anyone else has dealt with or considered this scenario?

backup rotation

3 Answers

  1. Best Answer

    If you're 1) trying to re-use tapes and 2) making sure that any file could be retrieved from any point in time, I posit that such an arrangement is impossible in the universe we live in.

    • 3

  2. I've seen what you're looking for, but only with to-disk backups in the mix. How such a rotation generally works:

    • Incrementals are spun to disk, daily.
    • The disk-version of the incrementals are spun to tape, generally multiplexed for fast processing.
    • The Incrementals are kept for a week.
    • Once a week, a "virtual full" backup is created.
      • A 'virtual full' takes the last Full (probably a virtual too), rolls the weeks Incrementals into itself, and publishes a new 'full' to the to-disk archive.
      • Once the virtual full is completed, it is spun to tape and archived.
      • The N-1 full is discarded.
    • The next Incremental is based off of that recent full.

    The advantage to the above is that your incremental size is a lot lower, so you'd consume fewer tapes for those. You are taking more fulls, so the difference may be a, er, wash in the end. The level of duplication in your on-tape backups allows you better recovery options, as you'd have once-a-week fulls to pull good data from rather than once-a-quarter.

    Additionally, if you can preserve your incremental tapes you can even get to day-before-corruption; the vastly shorter incremental chain means restoring that kind of data is much, much less painful.

    • 2

  3. I would argue that such a backup schema can not exist.

    The only thing that could help is if you had a method available to verify the integrity of all your files before throwing any backup away.

    • 0