On Fri, 9 May 2008, Marcus Herou wrote:
Ok so the mantra is: Backup as often as you can afford :)
I don't really see that as being a workable solution.
However if an index is corrupt it's not a big deal since the data is not connected to the index itself but rather stored elsewhere. Actually it will probably be stored with GlusterFS as well but with another access pattern which I hope will not run int these issues. This data will probably be rolled out on tape or such every night and a hot backup be kept somewhere just in case.
The problem is that hot backups will also be synced to the old version. Something like CopyFS on top of GlusterFS might give you what you need WRT backups, but I'm not sure how that would cope with versioning number resets and suchlike, and whether a broken sync would utterly break it.
Gordan
