I'm trying to find info on the error handling of the raid1 system. This
is an off-shoot of another thread looking for a backup archive format
with imbedded error correction.
All the raid documentation for linux talks about a drive either beeing
good or failed.
Suppose I have a 3-disk raid1 array and raid tries to read a block from
the first drive. If that returns an error, does it give up on the whole
drive and just use the other two? What happens if the next read it
tries to get a different block from the second drive and fails, yet that
block would be readable from the first drive if it tried?
I'm looking at setting up my off-line, off-site, backup strategy using
ruggedized external drives (USB or eSATA interface). If I decide to
have three sets of data to provide some data redundancy, could I set up
three external (ruggedized) drives in a raid1 then store them? Later,
if I try to mount one of the drives, if it gets an error, could I get a
second or third drive out of storage and put them back into the array
and have the raid1 subsystem make one intact md out of three damaged
Or would I be better to to use three drives non-raid then use some
recovery technique to generate the filesystem again. I'd rather have
such recovery be automatic and internal to the backup method.
I know that the drives themselves now use ECC so the errors I would get
would be if the drive wasn't able to regenerate the block.
I don't know what these errors look like since I haven't had a drive
with ECC fail yet.
I know someone will say to just backup to another computer but that
doesn't address the off-site nature. I can't backup via the internet
because I'm on dialup and the phone isn't reliable enought to stay up
long enough to transfer 80 GB (besides, I may want the phone for
something else). I can put two drives in the bank and mail one to my
Tape is too expensive and there's still be problem of handling read
errors after a long time in storage. CD/DVD are too small and are not