Solaris

Hello,

Can someone advise the proper procedure for replacing a mirrored disk in SVM. I have checked the docs and various websites but the procedure seems to vary. This is what I would do...

1. Remove the db replicas from the bad disk.
2. Detach it from the mirror
3. Clear it with metaclear.
4. Replace the Disk
5. Copy the label from the good disk using prtvtoc and ftmhard
6. Create new db replicas
7. Create all the metadevices again then attach them which should start the resync?

Is this all necessary or can I just do a metareplace -e ?

I hope this is roughly right, What im not sure about is the need to use cfgadm to deconfig the failed disk then configure the new on back in, is this dependant on what hardware you are running?

Also, is a reboot required for any of this?

Many Thanks.

Even if you use metareplace you will have to partition the disk first.

From your list you would have to do

1,4,5,6 then metareplace -e.

Thanks for the reply, however when I keep getting errors from metadevadm relating to invalid device information on the replaced disk. This has led me to believe that I need to configure the new disk to get a new DevID generated.

I would use the method you have listed in the original post. I never use metareplace.

It seems tedious however, it is a lot cleaner in the end.

I have sued both! The trick is to use 'metareplace' in case the entire disk goes bad!! That is faster and works okay

If only some slices are bad, then you need to follow the 'tedious' procedure that you mentioned in the very first post.

Extra:
Concerning usuing the cfgadm command, I have seen most sun engineers not using them while many sys admins use that - may be for security / satisfaction sake. Remember, cfgadm will eb used if you have scsi disks on your machine. ** 'luxadm' command is used instead of cfgadm incase your machine has fiber channel disks. (To find out if your system has FCAL disks, run format command and see if there is a long series of numbers after t (target) of the disks).

Hope that helps