From: jim kastner (jim.kastner@NWSCO.COM)
Date: Tue Jul 16 2002 - 06:28:54 EDT
This is our procedure. I hope this helps.
Replacing a Mirrored Database Disk Drive
1. Backup the database, then shut it down if it restarts after the
backup. DO NOT
ATTEMPT THIS PROCEDURE ON OPERATING DRIVES.
2. Check the error log using errpt -a to obtain the pdisk number of
the failed or
failing disk.
3. Use ssaxlate -l pdiskx (where x is the pdisk number from Step
2) to determine the
hdisk associated with the pdisk.
4. Use lspv -l hdisky (where y is the hdisk number from Step 3) to
determine the Logical
volume names contained on the defective disk.
5. Use lslv -m lvname (where lvname is the name of the Logical
Volume from Step 4)
to determine the hdisks that make up the mirrored Logical Volume
copies.
6. For the NON-DEFECTIVE hdisk in Step 5, use lspv hdisky to look for
stale
partitions. There should be 0.
7. Use rmlvcopy lvname 1 hdisky (where lvname is the name of the
Logical Volume
from Step 4 and where y is the hdisk number from Step 3) to break
the mirror.
8. Use reducevg ldbvg hdisky (where y is the hdisk number from Step
3) to remove
the defective disk from the ldbvg volume group.
9. Use rmdev -dl hdisky (where y is the hdisk number from Step 3)
to remove the
hdisk device. You always perform this step, even though it seems
counter-intuitive.
10. Use rmdev -dl pdiskx (where x is the pdisk number from Step
2) to remove the
pdisk device. You always perform this step, even though it seems
counter-intuitive.
11. Physically replace the drive.
12. Use cfgmgr to make the system recognize the new drive.
13. You should now be able to see the drive when you use lsdev -Cc
disk
14. Assuming the drive is recognized by the system, use extendvg ldbvg
hdisky
(where y is the hdisk number from Step 3) to add the new drive
back to the ldbvg
volume group.
15. Use mklvcopy lvname 2 hdisky (where lvname is the name of the
Logical
Volume from Step 4 and where y is the hdisk number from Step 3)
to recreate the
mirror copy structure.
16. NOTE: the following command must complete before starting the
database, and it
usually takes at least 45 minutes to run. Use syncvg -l
lvname (where lvname is
the name of the Logical Volume from Step 4) to resyncronize the
mirrors. It should
not be used with an actice database, as performance will suffer
terribly.
-----Original Message-----
From: Caldwell, Michelle [mailto:Michelle.Caldwell@AVNET.COM]
Sent: Monday, July 15, 2002 3:02 PM
To: aix-l@Princeton.EDU
Subject: replacing a mirrored disk for non-rootvg volume group
I am looking for a paper on IBM's website which tells me the simple
steps to recover from a failed disk with mirrored logical volumes
within a volume group. The volume group consists of 3 disks. All these
disks are attached to a SCSI raid controller which was setup with Raid 0
Stripping. So, what we did was stripe on hardware and mirror in AIX.
Can anyone help with a white paper or step by step instructions to
handled the mirrored lvm, reduce the vg, replace the disk for a
non-rootvg volume group?
Thank You
Michelle Caldwell
AVNET TECenter
Solutions Consultant
480-794-2743
This archive was generated by hypermail 2.1.7 : Wed Apr 09 2008 - 22:16:04 EDT