CFS/CVM Device Tree Cleanup Procedure

Problem

This procedure is to be used to cleanup /etc/vx/disk.info file when it shows a large number of devices having "0xffffffffffffffff 0x2" instead of the device hex serial numbers.

Error Message

In /etc/vx/disk.info file

HITACHI%5FOPEN-V%5F100D0%5F0002 sdd 0xffffffffffffffff 0x2 hitachi_r7000_0002 Hitachi_R700 65744
HITACHI%5FOPEN-V%5F100D0%5F0000 sdc 0xffffffffffffffff 0x2 hitachi_r7000_0000 Hitachi_R700 65744
HITACHI%5FOPEN-V%5F100D0%5F0006 sdf 0xffffffffffffffff 0x2 hitachi_r7000_0006 Hitachi_R700 65744
HITACHI%5FOPEN-V%5F100D0%5F0004 sde 0xffffffffffffffff 0x2 hitachi_r7000_0004 Hitachi_R700 65744
HITACHI%5FOPEN-V%5F100D0%5F012A sdaz 0xffffffffffffffff 0x2 hitachi_r7000_012a Hitachi_R700 65744
HITACHI%5FOPEN-V%5F100D0%5F012B sdcy 0xffffffffffffffff 0x2 hitachi_r7000_012b Hitachi_R700 65744
HITACHI%5FOPEN-V%5F100D0%5F0130 sdbf 0xffffffffffffffff 0x2 hitachi_r7000_0130 Hitachi_R700 65744

 

Solution

This procedure should be performed on the 'Slave' nodes in the cluster first.   Once it has been performed on all the 'Slave' nodes, then the 'Master' node is changed to become 'Slave' - and one of the 'Slave' nodes becomes the 'Master'.   The procedure can then be performed on the last node (was Master) in the cluster.

First determine which node is the "Master" node:

From lab testing:

# vxdctl -c mode

mode: enabled: cluster active - MASTER
master: sprnr3108f-22
 

Run the following commands on each of the Slave systems:

# rm /dev/vx/dmp/*

# rm /dev/vx/rdmp/*

# mv /etc/vx/disk.info /etc/vx/disk.info.OLD

# mv /etc/vx/array.info /etc/vx/array.info.OLD

# /sbin/vxconfigd –k

= = = = = = = = = = = = = = =

Now on Master system: 

The 'Master' needs to be changed to one of the 'Slave' systems before this procedure can be run on this last system.

In testing the master = sprnr3108f-22

Syntax:

# vxclustadm setmaster <system*>

* <system> = Slave hostname

Example:

# vxclustadm setmaster sprnr3108f-23

Check for master to verify that 'Master' has changed to another node: 

# vxdctl -c mode

mode: enabled: cluster active - SLAVE

master: sprnr3108f-23

Now perform the procedure on system that 'was Master' and is now 'Slave':

Note: In the lab test it was sprnr3108f-22

# rm /dev/vx/dmp/*

# rm /dev/vx/rdmp/*

# mv /etc/vx/disk.info /etc/vx/disk.info.OLD

# mv /etc/vx/array.info /etc/vx/array.info.OLD

# /sbin/vxconfigd –k

= = = = = = = = = = = = =

If wanted, the 'Master' can be switched back to the system that was Master.   This must be run from the current 'Master' node; in the lab test was sprnr3108f-23.

# /opt/VRTS/bin/vxclustadm setmaster sprnr3108f-22
 

And check:

# vxdctl -c mode
mode: enabled: cluster active - SLAVE
master: sprnr3108f-22
 

Applies To

5.1 SP1PR2 CFS/CVM running on RedHat 6.0.

This can apply to other platforms and product release versions running CFS/CVM cluster.

Note: Related Article points to the vxvm device tree cleanup.