[Henry J. Cobb]
> but if you've got a journaling filesystem, wouldn't you want to expose the
> raw disks to it so it can choose to put the journals on different disks
> than the files?
Funny, since sct/ext3 is the only one that appears to be pushing to keep
alive the possibility of journaling to other devices (nvram for one,
which is definitely a good idea). In one sense, creating an external
dependency for the recovering of your data can be a Bad Thing.
> This would not only help with performance, but it would also make recovery
> as simple as using one of the surviving journal copies and applying that
> against the last full backup of the main file system. (I.e. you lose 10
> disks out of your 12 disk "array" and wind up not losing a single byte of
> data.)
journals aren't *nearly* that deep. journal transaction entries can get
overwritten (circular buffer) as soon as the full transaction has been
committed to disk. It does *not* keep all transactions around since
your last full backup (how would it even know? :)
Journaling != RAID != LVM != Backups. They all serve their own purpose,
and invariably trying to use one to cover the tasks of others *will*
bite you eventually (as we have seen on this list multiple times)
James
