Replacing hard drive in FC10

[Dennis Kennedy]

Hello all hope thanksgiving went well.

I have to replace a hard drive in a FC10(fiber channel disk array). I have a plan set out to do this but it contains a "hope" that I don't like and can't rely on. I just want to bounce this off your heads.

Halt the system
Power off
Remove the old drive (Having read/write errors)
Insert the new drive
Hopefully it recognizes the hard drive as the old drives Physical volume. (Otherwise pvcreate then recreate the logical volume to point to this physical volume?)
Power on system
Restore data on the disk.

I have thought about dd but with the read/write errors is not probable. Two logical volumes are using this drive.

I have about a little over an hour window to do this so I have to be precise.

Any advise would be appreciated.
Thanx in advance
D

[Patrick Wallek]

Is this in a actual disk array that is using RAID 5 or is this set up stand-alone with no mirroring or anything?

If it is RAID 5 then the disk should be hot swappable and do an auto-rebuild of the array. If not, then your plan sounds good.

[Patrick Wallek]

Something else -- If it is not part of a raid 5 -- It probably won't see that drive as part of the volume group. You will need to do a vgcfgrestore (vgcfgrestore -n vgname /dev/rdsk/cXtXdX)

Do a man vgcfgrestore for more info.

[Dennis Kennedy]

Unfortunatly just a straight disk array, I probably should have added that. Raid would be great right now just swap and go. lol
Thanx anyway Patrick

[Patrick Wallek]

Without the Raid5 you are kind of stuck.

Your steps sound good in that case, with the addition of the vgcfgrestore in place of your pvcreate.

Hopefully you don't have much data to restore or have a really fast tape drive. :)

[Bill McNAMARA_1]

I never thought the FC10 could do RAID?!
I though it was just a JBOD.

In any case you don't need to power the
entire enclosure down just remove the failed
disk, replace the new one in the same slot.

The disk array will assign a loop id
dependant on the slot position and your
device file will be exactly the same.
Now, if you rebooted your host, the state
of NO_HW on the failed drive on an ioscan -fnk
output will vanish.. ie no hardware found.
You will have to run an ioscan -fn (or reboot)
to see the replacement disk and
insf -H8/12.8.0.255.0.7
to install the device file HW path is an example.

Then you must restore your LVM configuration
to the disk.
Find out what VG it's in.
strings /etc/lvmtab and look for device file.
then
vgcfgrestore vg
Because its a JBOD, unless you have mirroring
you'll have to restore data from backup.
a vgchange activation or vgsync will resync
your mirrors, If you're doing RAID? on it
I'm not sure how you'll restore it.. it may
be automatic, but if you're doing RAID, that
means you've got a lun and probably lvm has
nothing to do with the rebuild/restore.
For sure LVM doesn't do RAID 5.

Later,
Bill

PS don't forget the ESD strap. Those disks
are worth money!!!!

[Dennis Kennedy]

I will be replacing the disk tomorrow with a hp representative. 2 heads better than 1. I will document the steps and post them here. I will award points accordingly.
Thanx all for your input.
D

P.S. unfortunatly no raid on these disks

[Dennis Kennedy]

Hello All,

Here are the steps for a successful disk swap.

1. Made sure had good backups.
2. Shutdown the server.
3. The FC10 is hot swappable, no need to power it down. Swapped for identicle hard drives.
4. Powered on the server.
5. Ran "vgcfgrestore -n vg01 /dev/rdsk/cXtXdX"
6. "vgchange -a y /dev/vg01"
7. "newfs -F vxfs /dev/vg01/rlvolX - for each logical volume that had any data on the disk.
(pvdisplay -v /dev/dsk/cXtXdX |more).
8. Mounted the logica volumes
9. Restored each lvol affected.
10. :-)

Thanx all again
D