Solaris Volume Manager Drive Replacement

[kHz]

I have a production server that contains two internal SCSI drives: c0t0d0 and c0t1d0 that are mirrored using Solaris Volume Manager.

Another server is being ordered to add into the cluster. Can the second mirrored drive c0t1d0 be removed from the old server and placed into the first controller and target of the new server? Then take one of the new server’s internal drive and replace that in the old server and reestablish the mirror?

+--------+
|c0t0d0 |
| |
|c0t1d0 |
+--------+
Old Server
Take c0t1d0 and place this in “New Server” c0t0d0 and establish mirror on “New Server”

+-------+
|c0t0d0 |
| |
|c0t1d0 |
+-------+
New Server

Take c0t0d0 and place this in “Old Server” and reestablish the mirror on “Old Server”

Since there are only two drives it is an alternate boot disk. What steps on both sides would need to be taken?

Thanks.

 

[Annihilannic]

I can't think of any reason why you couldn't do that...

Are you familiar with the initial configuration of Solaris Volume Manager? Once you have swapped the disks you would have to copy the partition tables to the new disks (e.g. prtvtoc /dev/rdsk/c0t0d0s0 > /tmp/vtoc and fmthard -s /tmp/vtoc /dev/rdsk/c0t1d0s0), fix the state database replicas (because both servers will think that some of them have gone missing), and resync the mirrors.

It may be better to put the disk you move into the same slot on the new server to avoid metadb confusion (I don't know whether that would be a problem) and boot off it manually until you have mirrored it to c0t0d0s0.

Annihilannic.

 

[kHz]

Good idea about keeping it in the same slot on the new server.

On the old server after the drive is replaced, I would need to:

create identical partition table (prtvtoc, fmthard)
create state database replicas
run metareplace

Or am I missing a step or two?

On the old server:

same steps as above except I would boot from the new disk that was replaced.

Is this correct?

I have replaced failed drives before but have not taken a good drive from a good mirror from a production system and moved it to another server. I have used raidctl and the internal raid device to do the same thing I want to do with SVM and that works fine, but again, have not done this with SVM.

Thanks.

 

[Annihilannic]

I think you may have to delete the old state database replicas before creating new ones; even though they aren't really on the new disk, they will probably be listed in metadb -i with an F flag for 'replica has format problems', or something like that.

Otherwise I think you've got it... although in practice you frequently find there's more to these things!!

Annihilannic.

 

[kHz]

Thanks.