> > You're assuming the drive has a 2M cache. What if it has 1M, or 8M? ;) > > I'd like an option to specify such things. > > > Also - you're assuming that you can actually manage to force the disk to > > *write* the data to the disk surface *now*,
If you really need to repair the disk and are worried about this, there is usually a jumper on the hdd that can be set to disable the caching. Does horrendous things to performance, but it's designed for people who need to know that once it's gone to the disk, it's on the media, not running in memory. > That raises an interesting issue (something that was debated in SERVER-LINUX > recently). How do you construct a RAID type array to save your data if one > drive returns random junk? > > Currently it seems that the only option is a RAID-1 of three disks where the > OS reads from all disks (only available on AIX AFAIK). Use RAID 5 on your array, which has checksumming or parity strips built into every block saved to validate data. You need a minimum of 3 disks though. Regards, DaveN
