inconsistency /etc/lvmtab file in MC/SG

[chae, jeong-seong]

Hi~

/etc/lvmtab file of active node is different from standby node.
Standby node disappear the information of shared disk.
So Standby node don't work startup correctly.
In this instance, recreate the /etc/lvmtab file?

cluster package error log message.

########### Node "ktfsb": Starting package at Sun Sep 26 09:26:23 KST 2004 ###########
Sep 26 09:26:23 - "ktfsb": Activating volume group vg03 with exclusive option.
vgchange: Volume group "/dev/vg03" does not exist in the "/etc/lvmtab" file.
ERROR: Function activate_volume_group
ERROR: Failed to activate vg03
Sep 26 09:26:23 - Node "ktfsb": Deactivating volume group vg03
vgchange: Volume group "/dev/vg03" does not exist in the "/etc/lvmtab" file.
ERROR: Function deactivate_volume_group
ERROR: Failed to deactivate vg03
Sep 26 09:26:23 - Node "ktfsb": Deactivating volume group vg04
vgchange: Volume group "/dev/vg04" does not exist in the "/etc/lvmtab" file.
ERROR: Function deactivate_volume_group
ERROR: Failed to deactivate vg04
Sep 26 09:26:23 - Node "ktfsb": Deactivating volume group vg05
vgchange: Volume group "/dev/vg05" does not exist in the "/etc/lvmtab" file.
ERROR: Function deactivate_volume_group
ERROR: Failed to deactivate vg05

########### Node "ktfsb": Package start failed at Sun Sep 26 09:26:23 KST 2004 ###########

active node

# strings /etc/lvmtab
/dev/vg00
/dev/dsk/c2t0d0s2
/dev/dsk/c2t1d0s2
/dev/vg03
/dev/dsk/c5t0d1
/dev/dsk/c5t0d2
/dev/dsk/c7t0d1
/dev/dsk/c7t0d2
/dev/vg04
/dev/dsk/c5t0d3
/dev/dsk/c5t0d4
/dev/dsk/c5t0d5
/dev/dsk/c5t0d6
/dev/dsk/c5t0d7
/dev/dsk/c5t1d0
/dev/dsk/c5t1d1
/dev/dsk/c5t1d2
/dev/dsk/c5t1d3
/dev/dsk/c7t0d3
/dev/dsk/c7t0d4
/dev/dsk/c7t0d5
/dev/dsk/c7t0d6
/dev/dsk/c7t0d7
/dev/dsk/c7t1d0
/dev/dsk/c7t1d1
/dev/dsk/c7t1d2
/dev/dsk/c7t1d3
/dev/vg05
/dev/dsk/c5t1d4
/dev/dsk/c5t1d5
/dev/dsk/c7t1d4
/dev/dsk/c7t1d5

standby node
# strings /etc/lvmtab
/dev/vg00
/dev/dsk/c2t0d0s2
/dev/dsk/c2t1d0s2

Solve?

[Sridhar Bhaskarla · Accepted solution]

Hi,

YOu will need to create the map files on the active node and import the VGs on the standby node. I assume other than vg00, all the VGs are shared through serviceguard. Here are the commands to correct the issue.

1. Create map files and note the group minor numbers on the active node for each VG.

ll /dev/vg05/group (note the minor)
vgexport -p -v -s -m /tmp/vg05.map vg05

2. Copy /tmp/vg*.map files onto the standby node.

3. On the standby node import the VGs

mkdir /dev/vg05
mknod /dev/vg05/group c 64 0x050000 (replace 0x050000 with the one you noted in step 1)
vgimport -v -s -m /tmp/vg05.map vg05

Repeat the above for all the VGs. Do not activate the VGs. Try starting the package and it should work.

-Sri

[Shawn M Harris_1]

Using vgexport and vgimport with the -s (share) option simplifies the command but can have two cons depending on your hardware config.

The shared option saves the VGID in the map file and upon vgimport -s an attempt to scan all visible disks to find the VGID on those drives, taking much longer depending the number of drives available.

Second if you have two host bus adapters (which it looks as though you do) the preferred path for all LUNS will be down channel c5 and only use c7 channel for an alternate link. Potentially you could alternate the preferred path on the LUNS to load balance accross your hbas.

[Sundar_7]

On active node

# mkdir /root/MAPS
# for VG in vg03 vg04 vg05
do
vgexport -p -v -s -m /root/MAPS/$VG.map /dev/$VG
echo "$VG\tls -lrt /dev/$VG/group | awk '{print $6}'" >> /root/MAPS/groupfile
done

Copy the directory /root/MAPS to the other node

# rcp -pr activenode:/root/MAPS /root/MAPS
# cd /root/MAPS
# for VG in vg03 vg04 vg05
do
mkdir /dev/$VG
MINOR=$(grep "^$VG" groupfile | awk '{print $2}')
mknod /dev/$VG/group c 64 $MINOR
vgimport -v -s -m $VG.map /dev/$VG
done