vgreduce command

[Torsten.]

I see you have another SCSI HBA in slot 4 of the server (0/8/0/0.14.0) connected to controller B of the SC10.

So the failed disk is connected to controller A.

You are using high addressing mode, for this reason the failed disk must be in slot 0.

You did not post the full output of "ioscan -fn" and cstm, so I'm only "almost" sure.

[Fabio Ettore]

Hi,

just a clarification about vgreduce.
I know that it's not strictly necessary to vgreduce the disk before of replacing it. I suggested doing that just to do things cleaner and also to reply to Mark question:

- Is that good enough, or should I use the "-A" argument

I mean that when taking out the disk without vgreduce you will see ??? on PE for lvols mirrored and PV status unavailable by vgdisplay for that disk. This is sometimes more a bother than a real problem.
So replacing the mirrored disk without vgreduce is possible.

However good luck...

Best regards,
Fabio

[Torsten.]

Regarding "-A":

You can use

# lvreduce -m 0 -A n ...
and
# vgreduce -A n ...

Doing this will reduce the disk from the VG, but the config will NOT be safed.

After replacing the disk you can simply do a

vgcfgrestore ...
and
vgsync ...

because your config remain the same (mirrored).

[Emil Velez]

The "Managing systems and workgroups" manual
has detailed procedures for this


see

http://docs.hp.com/en/B2355-90950/index.html

0. make sure you have a good backup and ignite tape.
1. replace disk
2. ioscan
3. vgsync will work if disk same size
and same address
4. need to put back LVM boot structures
and of course test booting from disk afterwards