change a path of the VG of the cluster

[Jorge Prado T]

when failing a switch of the SAN, and after changing it, then the system change one of the path of the disks that compose the vg of the cluster, besides the cluster disk loock.
Which the procedure is to configure the VG of the cluster with the new path?
Which the procedure is to change the new path of the cluster disk lock?
S.O = Hp-ux 11.11
SG 11.16

Thank you in advance
Jorge Prado

[Steven E. Protter · Accepted solution]

Shalom again Jorge,

Your cluster no longer functions due to the absence of a lock disk.

One way to fix it is to re-run the cmcheckconf and cmapplyconf

In this way you can chose a new cluster lock disk.

You can also use sam if you are relatively unfamilair with cluster configuration or just feel like it.

Sam can help you pick a valid cluster lock disk.

Once you pick it, you change the cluster configuration file and cmcheckconf/cmapplyconf

Just like its a new cluster.

Package configuration will be preserved.

Re-start your cluster.

SEP

[Rajeev Shukla]

Ideally the path and device files should not change when changing a switch as long as the ports are maintained in the fabric. Meaning you connect the devices back to the same port on the switch after the swicth is replaced.

If the device files have changed the only way to reconstruct the VG with the new device file and new path is by doing vgimport using the map files of the VG.
Normally you should backup the vgexport map files (vgexport done with -s option). This helps in importing even if the device files or the path has changed.

Procedure to change the cluster lock disk.

First identify the cluster lock disk and then edit the cluster configuration file and change accordingly to reflect the new cluster lock disk.
do a cmcheckconf using the configuration file and then cmapplyconf

[YoungHwan, Ko]

If your system's H/W Path is modified you should use map file to recover apply new path.

# vgchange -a n vg##

# vgexport -p -v -s -m /tmp/vg##.map /dev/vg##

# vgimport -v -s -m /tmp/vg##.map /dev/vg## /dev/dsk/c#t#d0 /dev/dsk/c#t#d1 ...

# vgchange -a y vg##

Check /etc/lvmtab file

[Bharat Katkar]

HI Jorge,
If it's a small cluster and you won't spare downtime then shortest way would be to create soft link of the old device to new device.

But you need to remmeber this in future if any hardware changes happens.

e.g old path is
/dev/dsk/cxtxdx and /dev/rdsk/cxtxdx

After switch failure it might have become something like:
/dev/dsk/cytydy and /dev/rdsk/cytydy

Then simply create a link as:

# ln -s /dev/dsk/cytydy /dev/dsk/cxtxdx
# ln -s /dev/rdsk/cytydy /dev/rdsk/cxtxdx

DO it on all nodes of the cluster.

Regards,

[Mark Nieuwboer]

Hi Jorge,

The correct way is wat SEP was telling you.
Just rerun cmquerycl again and make the cluster again.

grtz. mark