Re:
> I have a disk drive going bad (I/O errors at around 6/minute). I'm calling
> HP to come replace the drive, but I would like to keep the process as
> efficient as possible. I have MPEX and DeFrag/X and I think I have enough
> space on the other drives to hold what's on that one. However, there is only
> one volume set. What steps am I best advised to take?
I discussed with Ted the pros/cons of cloning a failing disk,
and the alternatives (including ISL> DISCUTIL's SAVE command).
Your HP CE should be able to offline clone a disk by using
a utility to copy it to tape, replace the disk, and then
use the utility to copy the tape to disk.
<plug, somewhat>
Lund's De-Frag/X allows online (with MPE up) cloning of a disk drive,
with some caveats (e.g., have the system as idle as possible,
and if the source drive is in a mounted volume set, then
power cycle the computer as soon as the clone command is done ...
to minimize the chance that data on the source disk gets changed
by MPE after it is copied to the clone disk).
In any of the 3 above alternatives, you have to ask: how concerned
am I that I'll get bad data from the failing drive? Unfortunately,
I know of no way to always reliably answer that question.
If you're getting, say, an error from the same spot on the disk drive,
De-Frag/X's FINDSECTOR/FINDPAGE command will tell you what file(s) that
spot belongs to. HP has a password-protected utility to do the
same thing. In this case, you can decide whether or not to
revert to a prior backup of that file(s) after your recovery is
done.
If you're not sure where the errors are, and/or if you're getting
errors from lots of places, you may decide that the best bet is
to not risk copying data from the bad drive. In that case,
you would not want to do either form of clone. You could, however,
do an ISL> DISCUTIL, unmount the failing drive, and then SAVE all
files modified since the last backup ... this will skip any file
that's entirely or partially on the failing drive, however, so
you could end up with a mismash of today's data and yesterday's
data.
In short, the best way of insuring 100% chronological integrity
is to do a full restore from the last backup(s). Unfortunately,
that means either losing a day (or more) of data changes, unless
you were able to successfully save some kind of change log file
(perhaps via DISCUTIL/SAVE).
The advantage of cloning is that it's the fastest method of getting
your system back up. I recently used De-Frag/X to clone
a 4 GB disk in 105 minutes on a 3000/917 (single-ended SCSI).
The online clone technique requires:
connect a new disk drive (preferably the same size, if not, then
larger than the failing drive) to your HP 3000 at an unused
address.
configure it into the system, perhaps online via IOCONFIG
(check for old copy of LOG4ONLN.PUB.SYS first, possibly
purging it ... it may have old / obsolete / incorrect I/O changes
recorded in it)
idle the system...abort all batch jobs, shutdown the network,
LIMIT 0,0. Abort any daemons that you can (including the .../stm/...
processes on 6.5 and later) (SHOT can help you do this, of course :)
(The goal is to stop anything that might change a page on the
failing disk after the "clone" command is started!)
:run defragx.dfx.lps
clone <olddrive> <newdrive>
<power HP 3000 off... *not* =shutdown ... when done>
disconnect failing drive
bootup HP 3000.
deconfigure whichever ldev is now not in use (possibly the
failing drive's ldev, possibly the new one's ldev (if you
changed it's address during power off to match that of the
old drive)).
The offline clone is:
Use the ODE / COPYUTIL HP utility:
ISL> ODE
COPYUTIL
DISPMAP ... displays map of disks/tapes
BACKUP ... copies disk to tape
replace drive
RESTORE ... copies tape to disk
reboot
Then, lobby HP to provide a "CLONE" command in COPYUTIL, to
copy from disk to disk :)
Stan Sieler [log in to unmask]
www.allegro.com/sieler/wanted/index.html www.allegro.com/sieler
|
