Problem statement:
We are configuring a new ES40. We set up a 0+1 RAID set
with
six 9GB disks on a RAID Array 3000 controller with two
HSZ22s.
In Tru64 V4.0F, the disk configuration utility sees the raid
set as
one disk and shows the size as 8xxxMB which is the size of 1
disk.
It should show about 26xxxMB which is the size of 3 disks.
The HSZ22 sees the size of the raid set correctly. How can
I make
Tru64 see the correct size for the disk?
Thanks,
Mike Taylor
mike.taylor_at_scott.af.mil <mailto:mike.taylor_at_scott.af.mil>
NCI Information Systems
I received 4 replies so far that all said to clear the disk label and create
a new disk label.
I did the following and it fixed the problem. The disk configuration
utility now says the
size is 26xxxMB.
# disklabel -z rz24
# disklabel -wr rz24 hsz20
We possibly should have used hsz22 or ra3000 for the disktype. I am
not sure if
that would make any difference.
I include the replies because they contain some additional info that may
help others.
This is the first time I used this list. Thanks for all the help.
Mike Taylor
mike.taylor_at_scott.af.mil <mailto:mike.taylor_at_scott.af.mil>
NCI Information Systems
Replies:
Russ Fish
I may be off base here, but I just saw the same problem yesterday when I
added drives to a RAID install. When the drives came up, they were labeled
(somehow) with the wrong sizes. I deleted the disklabels with disklabel -z,
then re-disklabeled them with -wr, and the correct sizes appeared. Easy to
try and remove as a possible source of the problem...
Joe Ledesma
I prefer to use MAKEDEV, file, disklabel, mkfdmn, etc instead of the disk
configuration utility.
Are these disks also brand new? If they were used in some other array
previously, one of the disks may have an old disklabel on it (this is not
erased when the striped mirrorset is built).
In either case, try clearing out the disklabel (disklabel -z) and relabeling
it.
In the disk label, the total amount of sectors for the unit and for
partition c should be about 53307531 for a 6 member striped mirrorset using
9.1 GB drives (this number is actually from an HSZ70).
alan
Was one of the disks previously used a single JBOD or direct access
disk? What's probably happening is that the old disklabel is still present.
For that case, you can clear it and write a new one with:
# disklabel -z /dev/rrz#c
# disklabel -wr /dev/rrz#c ra3000
You might also want to check what the SCSI driver thinks the
capacity of the disk is. Use the "show capacity" command of scu(8) on the
device:
# scu -f /dev/rrz#c show capacity
If that's wrong, then a "scan edt" of the bus should fix it. Then
you can check the label and write a new one.
If the disk had a LSM sub-disk on it, LSM sometimes doesn't give up
devices easily. If LSM still thinks it has a sub-disk (plex or volume) on
that device, remove them with the appropriate LSM commands. If it still
think it is using the device, reboot. You shouldn't have to reboot to make
LSM forget, but sometimes it seems like the only choice.
Mandell Degerness
I would try writing a clean disk label to the disk.
* Note: the following assumes the disk is re0 - ie: disklabel -z
re0;disklabel -wr re0 SWXCR
This should cause the correct size to be written to the disklabel and all
should work correctly.
Received on Thu Jul 27 2000 - 19:06:20 NZST