Our Solaris 9 server has a disk c3t9d0 which reports via SMART that it
is failing. However, it has no uncorrected errors yet, so it should be
ok to just copy the data off of it (rather than restoring the whole
volume from tape). It is part of this lvm, from /etc/lvm/md.tab

d60 1 4 /dev/dsk/c4t0d0s2 \
/dev/dsk/c4t1d0s2 \
/dev/dsk/c3t8d0s2 \
/dev/dsk/c3t9d0s2

We have a spare disk which is exactly the same kind as these four.

It seems to me that the easiest thing to do would be to
1. shutdown (power off)
2. stuff in the new disk (for instance, as c4t2d0)
3. boot to single user
4. dd if=/dev/rdsk/c3t9d0s2 of=/dev/rdsk/c4t2d0s2 bs=2048
5. shut down again (power off)
6. pull the failing one
7. replace it with the new one
8. boot normally.

It looks like this copy also be done with "metareplace", but is there
any advantage in doing so? In terms of physically swapping the disk it
is in an external JBOD, and I'd prefer to power that (and therefore the
whole system) down when the disks are inserted or extracted. Down time
isn't a critical issue on this system.

Is there some other method which is better in some way than what I have
outlined above?

Thanks,

David Mathog