Replacing an AIX Fixed Disk

Overview

This document describes how to replace regular, stripped (Raid 0) and mirrored (Raid 1)
disk drives in AIX.

The procedure to replace an AIX disk depends on the situation. Considerations include
disk configuration, the type of volume group and logical volumes, and the extent of the
failure. Figure 1 in the appendix is a high level flowchart that helps identify the procedure
to use to replace the disk. The general disk replacement steps are:

1. Identify all LV's residing on the defective disk

2. Remove all physical partitions associated with the defective disk.
3. Remove the disk from its Volume Group and from AIX
4. Replace physical disk
5. Recreate LV's
6. Restore data

Standard Disclaimer

Please use this information with care. IBM will not be responsible for damages of any
kind resulting from its use. The use of this information is the sole responsibility of the
customer and depends on the customer's ability to evaluate and integrate this information
into the customer's operational environment.

Page 1 of 13
08/05/01
 Replacing an AIX Fixed Disk

Table of Contents

Overview ______________________________________________________________ 1
Standard Disclaimer _________________________________________________________ 1
Table of Contents ___________________________________________________________ 2
Replacing an AIX Fixed Disk _____________________________________________ 3
Before a Problem Occurs _____________________________________________________ 3
Special Case: Volume Group with One Disk _____________________________________ 3
Make a Backup _____________________________________________________________ 3
Identify All LV's Residing on the Defective Disk __________________________________ 4
Remove All Physical Partitions on the Defective Disk ______________________________ 4
Considerations for Removing System LV's (JFSLOG, Paging, Sysdump) _____________________ 5
Remove the Disk Definitions __________________________________________________ 6
Special Considerations _____________________________________________________________ 6
Physically Remove Disk ______________________________________________________ 6
Adding a New Drive to an Existing Volume Group ________________________________ 6
Recreate Logical Volumes and File systems ______________________________________ 7
Special Considerations _____________________________________________________________ 7
Appendix______________________________________________________________ 9
Document the Disk Configuration ______________________________________________ 9
Disk Replacement Strategies _________________________________________________ 12
Figure 1: Flow Chart for Replacing Disks ______________________________________ 13

Page 2 of 13
08/05/01
 Replacing an AIX Fixed Disk
Replacing an AIX Fixed Disk

Before a Problem Occurs

Document the disk configuration and make regular backups. Rebuilding a failed disk
requires knowledge of how it was configured. It may not be possible to retrieve the
configuration after a failure. For more information, see "Documenting the LV
Configuration" and "Making a Backup" in the Appendix.

Special Case: Volume Group with One Disk

If the defective drive is the only disk in the Volume Group, then use the three step
procedure:

1. exportvg <Vgname>
2. rmdev -dl <hdisk#>
3. restore from backup

Make a Backup

There are two general types of backup in AIX: one creates an install tape for the
operating system and the other saves data.

If you are having a problem with an operating system disk (rootvg volume group), you
need to reinstall your system. This requires a system backup (install image), which
requires special backup commands such as mksysb, alt_disk_install, or equivalent. The
mksysb command (smit mksysb) is standard method for making a system backup to tape
or CDROM. You can also use the alt_disk_install method to dynamically copy the
operating system to another disk, then reboot on the new disk. These commands are
limited to backing up the operating system (rootvg). They can not be used to backup data
in non-rootvg volume groups. Both commands are part of the base AIX operating
system

There are a number of commands to save data outside rootvg. They include tar, cpio, pax,
backup, dd, savevg. In addition, many applications and databases have their own backup
commands. Be sure to backup raw partitions as well as file systems. See the man pages
for more information.

The backup frequency depends on value of the data and its volatility. The more valuable
or volatile the data, the more it should be backed up. In addition, the backup quality
should be verified. Tapes can become defective or the backup procedure may have a
problem. Verify the backup by restoring to a different system, or by running a "table of
contents." Use a group of backup tapes, and use a different tape for each backup. If a
tape fails, you can use the previous backup tape.

Page 3 of 13
08/05/01
 Replacing an AIX Fixed Disk

Identify All LV's Residing on the Defective Disk

To list the LV's residing on a disk, run:

lspv -l <hdisk#>
-or-
lspv -l <PVID>

Sometimes the hdisk# becomes invalid. If so, try substituting the PVID for hdisk# (from
the lsps command).

Remove All Physical Partitions on the Defective Disk

All physical partitions (PP) must be removed from the disk to be replaced. The way to
remove the partitions depends on whether the disk is mirrored.

The first step is to stop applications and users from using the disk. If "normal"
commands don't work, use the fuser command as a last resort to kill all processes using
the logical volume.

fuser -kxuc /dev/<LVname>

On mirrored disks, remove partitions associated with each logical volume using the
rmlvcopy command. The command syntax is:

rmlvcopy <LVname> 1 <hdisk#>

<hdisk#> refers to the name of the failed disk. If the hdisk# is unusable, use the PVID.
The <LVname> is the logical volume name. Run this command for each LV on the
failed disk. In some cases the unmirrorvg command may be used in place of rmlvcopy.
See the man pages for each command to determine which command best fits your
situation.

On unmirrored disks, the entire LV must be removed, even if just one of its physical
partitions is on the disk to be replaced. The command syntax to remove a logical volume
is

rmlv <LVname>

If the LV contains a JFS file system, unmount the file system first. Then use the rmfs
command to remove both filesystem and logical volume at the same time.

Page 4 of 13
08/05/01
 Replacing an AIX Fixed Disk

umount /<FSname>
rmfs /<fsname>

Removing Special LV's

Page Space: Determine if an unmirrored paging space is active.

lsps -a

If it is active, set it to be inactive on the next reboot.

chps -a n <LVname>
Reboot to deactivate page space

Remove the inactive page space:

rmps <LVname>

JFS Log: If the unmirrored disk contains the JFSLOG (typically named loglvnn), you
won't be able to write to any files system in the Volume Group. To remove the JFSLOG,
unmount all file systems in the Volume Group, then use the rmlv command to remove it.
Record the name of the JFSLOG because you'll want to use the same name when you
recreate it.

System Dump Space: If the LV is serving as a dump device, the dump pointer must first
be reassigned. The same is true if the LV was mirrored and the copy is being removed.
Check the dump pointers by entering:

sysdumpdev -l

Reassign the dump pointers. Enter:

sysdumpdev -Pp /dev/sysdumpnull (for the primary device)

sysdumpdev -Ps /dev/sysdumpnull (for the secondary device)

Disk is Defective, But Not Failed: If the disk has not failed and there is another disk
available, you can minimize the repair work with migratepv command. This command
can be used to dynamically move the LVs off the failing disk to another. See man pages
for more information on this command.

Page 5 of 13
08/05/01
 Replacing an AIX Fixed Disk
Remove the Disk Definitions

Removing the disk definition is a two step process. The disk must first be removed from
its Volume Group, then remove the disk physical definition from AIX. Assuming we
want to remove hdisk10 from the datavg Volume Group, the commands would be:

reducevg -f datavg hdisk10

rmdev -dl hdisk10

If the hdisk definition is no longer valid, try replacing with its PVID (found using the
lspv command). You must also ensure that the PVID is removed from the ODM with the
following command. The 32-digit value supplied consists of the PVID plus 16 zeros. For
example:

odmdelete -q value=0073659c2c6d26f10000000000000000 -o CuAt

Special Considerations

SSA Disk: SSA disks also have a pdisk definition, which must be removed. The
association between hdisk and pdisk number is random. See "smit ssa" to determine the
pdisk number associated with the hdisk::

rmdev -dl pdisknn

Physically Remove Disk

To physically remove the hard disk, consult the documentation for that device, or the
hardware service organization for the vendor.

In general, "hot swap" and SSA drives may be replaced while the other disks on the bus
are active. Otherwise, do not attempt to remove a disk while other disks are active on the
same bus. Removing a disk can cause the SCSI adapter to reset, causing operating system
misdiagnose all disks on the bus as failed.

Adding a New Drive to an Existing Volume Group

If the failure was an unmirrored rootvg disk, reinstall the operating system on the new
disk. This is normally done from a "mksysb" image (tape or CD) or AIX install CD's.
See the documentation for installing AIX for more information.

Page 6 of 13
08/05/01
 Replacing an AIX Fixed Disk
Otherwise run the cfgmgr command to identify the new disk to AIX. This step is not
necessary if the system has been rebooted, as cfgmgr is run as part of the boot process.

Once the new drive has been identified, ensure that a proper PVID has been written to the
drive by running:

chdev -l <hdisk#> -a pv=clear

chdev -l <hdisk#> -a pv=yes

NOTE: On SSA drives, the first chdev command may be omitted.

Add the drive to the volume group:

extendvg VGname hdisk#

If this was the only drive in the Volume Group, use the mkvg command to recreate the
volume group on the new drive.

Recreate Logical Volumes and File systems

New logical volumes, paging spaces, file systems, or logical volume copies can be re-
added with one or more of the following with the mklv, mkps, crfs, mklvcopy,
mirrorvg, restorevg commands. All of these commands can be run through "smit".

If the failed disk was unmirrored, create new LVs using the mklv command. If you have
the map file, you can use the "-m" flag to recreate the partitions in the exact order and
physical location on the disk (see section on Documenting the Disk Configuration) . If the
LV contained a file system, use "smit crfs" to "Add a Journaled File Systems on a
Previously Defined Logical Volume".

If the failed disk was mirrored, use the "mklvcopy" command. Again, you can use a
MapFile to restore the copies to the exact location as before. After recreating the LV
mirror copies, be sure to synchronize them using the "syncvg" command. For example,
the "syncvg -P 6 -v <VGname>" synchronizes the entire volume group. See the man
pages for other options..

Special Considerations

Boot Disk: If the failed disk was a mirrored rootvg disk, be sure it is bootable. Verify hd5
is properly mirrored on the new disk. If one does not exist, see the bosboot
documentation for creating one. Define the disk as bootable to AIX For example, if
hdisk0 and hdisk1 are mirrored disks in rootvg, define them as bootable via the
command.

Page 7 of 13
08/05/01
 Replacing an AIX Fixed Disk
bootlist -m normal hdisk0 hdisk1

JFSLOG: This is a special partition for logging JFS file system changes. There is at
least one JFSLOG in every Volume Group containing a JFS file system. To recreate the
JFSLOG, first unmount all JFS file systems in its volume group. Delete and recreate the
JFSLOG on a disk in the same volume group using the same LV name and type JFSLOG.
Format the partition. Mount the file systems.

For example, assume the log file name is loglv01 and the volume group name is datavg,
and the target disk is hdisk10. The commands to recreate the JFSLOG would be:

# unmount all JFS file systems datavg

rmlv /dev/loglv01
mklv -y loglv01 -t JFSLOG datavg hdisk10
logform loglv01
# remount file systems in datavg

Bruce Spencer
IBM
baspence@us.ibm.com

Page 8 of 13
08/05/01
 Replacing an AIX Fixed Disk

Appendix

Document the Disk Configuration

I generally use these commands to document the disk configuration:

1. lspv
2. lsvg -l <VGname>
3. lspv -l <hdisk#>
4. lslv <LVname>
5. lslv -m <LVname>

1 Use the lspv command to identify the Volume Group containing the defective disk.
The output will list the hdisk, it PVID and the volume group to which it belongs:

# lspv
hdisk0 00000036960cbdd1 rootvg
hdisk1 00000036960c31de rootvg
hdisk2 00000036960d3007 vg01
hdisk3 000003870001328f vg01
hdisk4 00000360ebf34660 vg01
hdisk5 00000360d7c1f19f vg01
hdisk6 00000036628b9724 vg02

2. Use the lsvg command to list the Logical Volumes in the Volume Group. For example,
the output for vg01 might look like:

# lsvg -l vg01

vg01:
LV NAME TYPE LPs PPs PVs LV STATE MOUNT POINT
loglv01 JFSLOG 1 1 1 open/syncd N/A
lv04 jfs 31 31 1 open/syncd /vol/x11r4
lv27 jfs 53 53 1 open/syncd /vol/X11R5
lv08 jfs 13 26 1 open/syncd /vol/tex
lv10 jfs 63 63 1 open/syncd /vol/abc
lv21 jfs 84 84 2 open/syncd /var/spool/news
lv12 jfs 99 99 1 open/syncd /vol/lpp/3005
lv23 jfs 66 66 2 open/syncd /vol/src
lv07 paging 92 92 1 open/syncd N/A

This output tells us several things. First, it tells whether the logical volumes are mirrored.
In this case, only lv98 is mirrored (PP's = 2 x LP's). Second, it tells us the size of the
LV's if we have to recreate them (LP's). Third, it shows that lv21 and lv23 reside on 2
disks (PV=2).

3. Use the lspv -l command to list the LV's that reside on the defective disk.

Page 9 of 13
08/05/01
 Replacing an AIX Fixed Disk

# lspv -l hdisk3

hdisk3:
LV NAME LPs PPs DISTRIBUTION MOUNT POINT
lv08 26 26 02..10..10..04..00 /vol/tex
lv21 16 16 02..00..12..00..02 /var/spool/news
lv04 31 31 01..08..05..15..02 /vol/x11r4

If the hdisk name no longer exists, and the disk is identifiable only by its 16-digit PVID,
substitute the PVID for the disk name (see lspv). For example, hdisk3 would be

# lspv -l 000003870001328f

If you receive the following error:

0516-320 : Physical volume 000003870001328f is not assigned to a volume group.

Run the following command:

putlvodm -p `getlvodm -v <VGname>` <PVID>

For hdisk3 in vg01 type:

putlvodm -p `getlvodm -v vg01` 000003870001328f

4. The lslv command to lists configuration information for each LV on the defective
disk. Check for non-default settings in case they need to be recreated.

# lslv lv04
LOGICAL VOLUME: lv04 VOLUME GROUP: rootvg
LV IDENTIFIER: 0000059293ce0140.9 PERMISSION: read/write
VG STATE: active/complete LV STATE: opened/syncd
TYPE: jfs WRITE VERIFY: off
MAX LPs: 512 PP SIZE: 4 megabyte(s)
COPIES: 1 SCHED POLICY: parallel
LPs: 31 PPs: 31
STALE PPs: 0 BB POLICY: relocatable
INTER-POLICY: minimum RELOCATABLE: yes
INTRA-POLICY: middle UPPER BOUND: 32
MOUNT POINT: /vol/x11r4 LABEL: /vol/x11r4
MIRROR WRITE CONSISTENCY: on
EACH LP COPY ON A SEPARATE PV ?: yes

5. Optionally, use the "lslv -m <LVname>" command to map the physical location of
each partition for each LV. Convert the output to a Map File format, and use it with
the "mklv -m MapFile" or "mklvcopy -m MapFile" to recreate the partitions in the
exact order and physical location. For example:

Page 10 of 13
08/05/01
 Replacing an AIX Fixed Disk

#lslv -m lv04
LP PP1 PV1 PP2 PV2 PP3 PV3
0001 0016 hdisk2
0002 0017 hdisk2
0003 0018 hdisk2
0004 0019 hdisk2
0005 0020 hdisk2
0006 0021 hdisk2
0007 0022 hdisk2
0008 0023 hdisk2

The Map File format is "PV:PP". The following "awk" command is one way of
converting the "lslv -m"output to the Map File format. The example creates a Map File
for the first LV copy (columns 2 and 3).

# lslv -m lv04 | awk ' { print $3 : $2 } '

PV1:PP1
hdisk2:0016
hdisk2:0017
hdisk2:0018
hdisk2:0019
hdisk2:0020
hdisk2:0021
hdisk2:0022
hdisk2:0023

To create a map file for the second or third copy, simply change the column numbers I
the "awk" print statement

Page 11 of 13
08/05/01
 Replacing an AIX Fixed Disk

Disk Replacement Strategies

Table 1: Disk Replacement Strategies

Volume Group Unmirrored Disk Mirrored Disk
rootvg Backup: mksysb Backup: mksysb

Procedure: Procedure:
reinstall AIX Unmirror LVs residing
on failed disk
Remove/replace disk
Recreate mirrors

Non-rootvg Backup: Backup:

tar, cpio, backup, pax, tar, cpio, backup, pax,
etc etc

Procedure: Procedure
Remove all PP's on Remove all LV copies
failed disk on failed disk.
Remove/replace disk Remove/replace disk
Recreate LV's Recreate LV copies
Restore from backup

Page 12 of 13
08/05/01
 Replacing an AIX Fixed Disk

Figure 1: Disk Replacement Flowchart

Page 13 of 13
08/05/01