This question is very important to anyone with a lot of disk space on an
IBM RS6000 running AIX. I have already posted it to comp.unix.aix and
received a rather poor response. Since many AFS sites meet the above
qualification, I am posting it here.
Question: If one disk in a volume group fails, is there any way to
recover the logical volumes in the volume group which DO NOT have blocks
on the broken disk?
Specifically, the disks on one of my servers are organized as follows:
volume group:
disk size logical volumes (filesystems)
rootvg:
hdisk0: 203 hd1, hd2, hd3, hd4, hd5, hd6, hd7, hd8 (/, /usr, /tmp, etc.)
uservg:
hdisk1: 203 lv03
hdisk2: 203 lv02
hdisk3: 326 lv00
hdisk4: 326 lv01
hdisk5: 326 lv00, lv01
hdisk6: 326 lv02
(disk sizes are in units of 'physical partitions' (4096 x 1024b blocks))
The advantage of this arrangement is that lv00, lv01, and lv02 all are
the maximum possible size for a filesystem (500pp or 2GB). lv03 is 199pp
which can be extended to 500pp onto another disk when one is purchased.
Having maximum size filesystems minimizes free space fragmentation.
Theoretically, my exposure is limited to at most two user filesystems if any
particular disk fails.
Practically, I have been told that it is not possible to recover any
filesystems in a volume group which has had a disk fail.
Before I spend a LOT of effort reorganizing my disks and filesystems in
a less efficient configuration so as to avoid a horrible disaster, I would
like to be sure that it's absolutely necessary. My definition of a
horrible disaster is taking two days to restore 10GB because one 857MB disk
failed.
I would be most interested in the experiences, successful or otherwise,
of anyone who has actually tried to recover from such a situation.
Thank you.
-Rick
|Rick Cochran 607-255-7223 [EMAIL PROTECTED]|
|Cornell Materials Science Center [EMAIL PROTECTED]|
|E20 Clark Hall, Ithaca, N.Y. 14853 cornell!msc.cornell.edu!rick|
