Fibre Channel controller failed, how to update volume group disk definitions

[dictum9]

Have a FC controlled that failed. Now none of the volume groups activate other than VG00.

Running autopath and it's showing disks that are different from the disks that are used by LVM.

It used to show 4 paths, now only showing 2. The paths that are not being shown are the ones defined in the volume groups.

strings /etc/lvmtab

/dev/vg01_eva
/dev/dsk/c0t0d1

. . .

. ..

pvdisplay -v /dev/dsk/c0t0d1
pvdisplay: Warning: couldn't query physical volume "/dev/dsk/c0t0d1":
The specified path does not correspond to physical volume attached to
this volume group
pvdisplay: Warning: couldn't query all of the physical volumes.
pvdisplay: Couldn't retrieve the names of the physical volumes
belonging to volume group "/dev/vg01_eva".
pvdisplay: Cannot display physical volume "/dev/dsk/c0t0d1".
#

That same disk shows up as c2t0d1 and c7t0d1 in autopath display. How do I update the VG info with the correct link?

c0t0d1 is the same disk as c2t0d1 but the FC controlled failed, which causes none of the volume groups to activate.

I tried pvchange -a n /dev/dsk/c0t0d1 but could not run it due to volume group not activated, and unable to activate it.

I realize I will have to update the WWW name stuff once I get the FC controlled replaced, but I would like to bring up the VGs now via alternate links.

# autopath display | head -66

==================================================================
 HPswsp Version        : A.3.0F.03F.00F
==================================================================
 Auto Discover         : ON
==================================================================
 Array Type            : EVA-AA
 Array WWN             : 5001-4380-025C-3450
 Read Optimize         : ON
 Path Verification Period : 00:10
==================================================================
 Lun WWN               : 6001-4380-024D-0AFC-0000-6000-0094-0000
 Virtual Device File   : /hpap/dsk/hpap1
 Load Balancing Policy : Shortest Service Time
 Lun Timeout           : Infinite Retry (-1)
==================================================================
 Device Path                    Status
==================================================================
 /dev/dsk/c2t0d1                Active
 /dev/dsk/c7t0d1                Active

Tried this to no avail:

# pvchange -s /dev/dsk/c2t0d1
pvchange: Couldn't find the volume group to which
physical volume "/dev/dsk/c2t0d1" belongs.
#

[Bill Hassell]

You can export the VGs, then import the VGs.

If you have run vgexport -s -p -m <mapfile> vg_name on all the VGs, you'll have a MUCH EASIER time importing the VGs.

Without a mapfile, vgimport will require identifying disk LUNs that belong to specific VGs, then renaming the lvols to their correct names.

sysadmin note: EVERY HP-UX sysadmin must run the above vgexport coomand to create mapfiles for all VGs.

[sapoguheman]

What is the version you are runninng is it 11.23 ?

IT COULD BE THAT , YOU HAVENT USED ALL ALTERNATE PATHS OF THE DISKS IF YOU ARE RUNNING ON 11.23.

[dictum9]

11.23

[Torsten.]

All LUNs should be accessible via the remaining paths.

However, if you swap the FC hba in a server, you get a different WWN. So you need to adjust the switch zoning and the EVA LUN presentation, because all is based on the old WWN.

On both (switch and EVA) change the WWN-alias relationship to assign the new WWN to the existing host alias and it will work again.

[dictum9]

I exported to a map file, what is the next step?   vgexport then vgimport?

# vgexport -s -p -m /var/tmp/vg01_eva.map vg01_eva
# ll /var/tmp/vg01_eva.map
-rw-r--r--   1 root       sys             30 Jul 13 07:01 /var/tmp/vg01_eva.map

# cat  /var/tmp/vg01_eva.map
VGID 05fb4fc74a816535
1 lvol1
#

[dictum9]

> All LUNs should be accessible via the remaining paths.

I understand what you are saying and I will work on it.

However,  I can see half the paths. How do I configure the VGs to use these alternate paths? And I replace the FibreChannel card, use that too as a redundant path?

[Torsten.]

Because of the changed WWN of your FC HBA the switch and the EVA array will deny access to the LUNs, hence you have only half of the pathes working.

Do not change the LVM configuration!

Adjust the switch and EVA configuration!

Check if the switch is using WWN based zoning and add the new WWN to the host aliases. If this is port based zoning, do nothing.

Go to CV EVA and check the host section. Add the new WWN of the server to the host. That's all.

[dictum9]

The problem was, the volume groups only specified one path, one disk, one controller and it failed.

I recreated the VGs using all available controllers. I used vgexport and vgimport. It worked.

I followed this somewhat closely, except I added disks during the vgimport command.

http://wiki-ux.info/wiki/How_to_change_the_name_of_a_volume_group