The internals of software RAID volumes do not lend themselves
to make
determinations that a corrupted or degraded RAID volume is
"broken"
or of the old flavor.
Dan,
That puzzled me, as it is dead easy to build a version identifier or
something similar into something like a RAID header, so I checked
it up -
and sure enough Apple *does* build a version number into the
header, which
changed from AppleRAID 1 (pre-Tiger) to 2 (Tiger). So it would
appear to be
merely a matter of checking the version number - hardly onerous?
And since a corrupted disk may have headers which are unreadable...
It may, sure, but it didn't in this case. If I couldn't access 100%
of the filesystem that I was trying to repair without errors, I
wouldn't expect to be able to convert it without errors either. But
that wasn't the case here: apart from the failure of one disk (i.e.
half of the mirror), my filesystem was functioning perfectly until I
had to run diskutil to convert from AppleRAID 1 to 2. I don't see why
I should have been worried about that in advance, when I try and
upgrade something that's working I don't expect it to break in the
process.
"may" sure. But to date I have not heard of anyone who has
successfully
rebuilt a pre-Tiger AppleRAID from Tiger - and I simply don't
believe that
the headers were unusably trashed in all those cases. Rather it
looks like
the rule is that Tiger *cannot* rebuild pre-Tiger AppleRAIDs,
regardless of
the header state - iow no attempt is made to read the version
number and act
appropriately, either by having a pre-Tiger rebuild routine, or by
bailing
out and saying that it is AppleRAID v 1.
It has been a few months since I've done it, but I'm pretty sure that
the last time I tried this -- i.e. having an AppleRAID 1 array become
marked as degraded in Tiger, then convert & repair the mirror, it
worked fine. It just didn't this time; the worrying thing is from
comments like yours when you indicate it virtually never works in the
real world.
Of course this compounded by the lack of docs to clarify the nature
of the
problem.
