7

What is the general best practice for keeping LTO 5 tapes in a rotation?

I have read that they have archival capability of 30 years, but I am sure this really depends on how much you're reading and writing to them during their service period, how they are stored, etc. Obviously we aren't going to be using them for 30 years here, but I seeem to encounter errors on them more frequently as they get older (2+ years), with no particular pattern. I've heard some IT folks say 2 years in a weekly rotation is about all you should expect to get out of them.

Should tapes be taken out of service after

  • N number of years?
  • N number of mounts?
  • N Terabytes written to and read from?
  • Any other type of metric?
JoshODBrown
  • 335
  • 4
  • 13
  • The more important question is why are you even using tapes for backup anymore? We switched to backups over the network to our DR / co-location and never looked back, its all fully automated, and we don't have to mess with tapes. – MDT Guy Oct 16 '13 at 20:44
  • @MDTGuy Because tape is cheaper than disk in many ways, and the way backup vendors price tape-coverage vs disk-backup coverage incentivises tape use. $OldJob is undoubtedly still using tape to backup almost everything because they had such a hard time getting budget for a large enough stack of disk to do it that way. – sysadmin1138 Oct 16 '13 at 20:51
  • Let's not rehash that argument here, it's already been exhaustively considered at http://serverfault.com/questions/550763/what-are-the-advantages-of-tape-drives . – MadHatter Sep 08 '15 at 10:10

2 Answers2

9

Tapes intended for long term archive (over a year) should be used as few times as possible before going into that archive. Fresh tapes are much more likely to last the full 30 years than a tape that was pounded weekly for a year before getting put into the Annual Archive set.

But I think you're asking about service-life for actively used tapes.

A LOT of it depends on how your tape drives work during operation. The following factors will limit service-life:

  • How many full-tape accesses happen.
  • The speed that the tapes are written to.
    • If the backup speed is very stuttery, the drive will stop and start a lot, and may have to shift to different throughput rates a lot. This puts a lot of stress on the media.
    • If the backup speed is constant (say, streaming directly from the backup-to-disk set as part of a media rotation scheme), they'll last longer.
  • Mount count plays a role here, though that's less important than the previous point.
  • Environmental factors.
    • Store them in a climate-controlled room, even the offline archives.
    • Allow the media to acclimate before use if they're coming from a space with different temp/humidity settings.

Unfortunately there is no algebraic formula to figure out chance of tape failure based on numbers of things. You'll have to build a heuristic based on your own backup patterns.

sysadmin1138
  • 131,083
  • 18
  • 173
  • 296
1

My rule of thumb is to let my backup software tell me when to replace a tape in any of the regularly-used pools. All good software does verification passes and/or checks the status of each write block with the drive. For example, when my bacula shows me a line in bconsole's list volumes output that says, eg,

|      46 | A00035L5   | Error     |       1 | 1,746,844,710,912 |    1,761 |    1,209,600 |       1 |    6 |         1 | LTO-5     | 2015-09-08 00:26:24 |

I know it's time to replace that volume.

MadHatter
  • 78,442
  • 20
  • 178
  • 229