In ioscan device file got changed

[Arunkumar.B]

HI,

we're having cluster RAC configured in superdome connected to eva 4k.Device file for the disk got changed for two path without notice.Now how will i change the entries in vg.Now the server is working fine in production.Whether the vgreduce for the old device file & vgextend for the new device can be done in fly.Is there is any changes need to cluster configuration.


I am attaching the ioscan & strings o/p

[Albert_31]

Hello Arunkumar,

A change in the SAN has changed the address in the HPUX io table.

Regarding the solution.

This is a RAC setup right and from my understanding both the nodes will be accessing the vg in parallel (somewhat). All you have to do is a vgreduce and vgextend on the paths since they are all alternate paths to the same luns. So the VG will work till we have one path available to the luns.

No other changes are required, however ensure that both the nodes reflect the current situation.

regards

Albert

[Mustafa Gulercan]

if the server is working fine in production,
do not do anything ;)

good luck
mustafa

[Arunkumar.B]

Hi Albert,


Sorry to forget this line in query.

Total 2 Packages are running

one RAC Package (Active - Active) vg01

one More package (Active - Passive) vg02


Then for the second one i will have to export import once again in the servers

I have not done any activities in SAN. where we can find the logs for this ????

[Albert_31]

one RAC Package (Active - Active) vg01
Frankly never done the activity on this setups.. but looking at the way it works, vgreduce followed by vgextend should be the only way..else wait till you get downtime, since only the alternate paths are wrong..you still have 2 more paths. Anyways wait for more inputs from the forum.

one More package (Active - Passive) vg02
a) Active node
backup /etc/lvmtab
vgreduce (old paths)
vgextend (new paths)
vgcfgbackup
vgexport -p-v-s-m
copy the file on the other server

b) Passive Node
backup of /etc/lvmtab
vgexport -v -s -m
mkdir
mknod
vgimport -v -s -m (it will warn about cpu id ...., ignore it)
Note we have to take the vgcfgbackup of this vg on this node, which is not possible now.. but remember this and ensure you do it..

c) There is no logging on this on the OS but on the switch / storage..

[Arunkumar.B]

Hi Gurus,


AnyMore Updates ???

Cheers
Arunkumar.B

[BPatrick]

You can do a export and import from one node to the other. The device file names get changed when there is any change in the SAN fabric. Make sure to backup your lvmtab file before import. In the cluster you need not make any changes unless your cluster lock disk device name has changed.
You can check the EVA logs from the command view EVA.

[Enrico P.]

Hi,

for the vg02 the steps should be the follows:

In the node that have the vg02 active:

1)vgreduce -f /dev/vg02

This will remove all the c20 and c16 path from vg02.

You can remove one at once also with the

vgreduce -l vg_name pv_path
2) vgextend /dev/vg02 /dev/dsk/c23t0d3 /dev/dsk/c23t0d4 /dev/dsk/c24t0d3 /dev/dsk/c24t0d4
3) vgexport -v -s -p -m vg02.map vg02
4) copy the vg02.map file in the other node

In the other node:

1)ll /dev/vg02/group (sign the minor number 0xXX0000)
2) vgexport vg02
3) mkdir /dev/vg02
4) mknod /dev/vg02/group c 64 0xXX0000
5) vgimport -v -s -m vg02.map vg02

For the vg01 the step should be the same but I don' t think if the vg is active in shared mode you can run vgreduce and vgextend command. If not you need deactivated it with vgchange -a n, then with -S n, reactivate it in exclusive mode an follow the step.
Then you can reactivate it in shared mode.

Check if the lock disk special file changed, if so you need to change it in cluster ascii file and recreate the binary cluster file with the cmapplyconf command with cluster down.

For the log it should appear in the syslog.log file.

Enrico

[Arunkumar.B]

HI Guru's


I am attaching the file which is giving when i try to reduce the vg...



With cheers
Arunkumar.B

[Marvin Strong]

The volume group has to be active to do the vgreduce. Stop the package but leave the do not cmhaltnode. Then you can activate the vg in exclusive mode and vg reduce the volumes.

cmhaltpkg
vgchange -a e
vgreduce
vgextend

Then create the maps and move them to the other node.

[freddy_21]

is your RAC cluster using raw device or LVM.
If using raw device ithing with changes the device can make your RAC not working, you must configure back the disk device. Or you must changes at your RAC.

I have one customer have Cluster RAC, the device changes, so he need me to make a same configuration with existing. Or he will changes all configuration at the RAC. and take time for do that.

If your configuration using raw device for disk, please backup your ioconfig, if the device changes you can rollback to existing configuration

thanks
Freddy

Thanks
Freddy

[Arunkumar.B]

Hi Guru's


There is two way i can do..Guide me what should i proceed.Primary links are fine.

After halting the cluster (It's not working)

1)vgreduce the old device files
2)vgextend with the new device files


or
1)mv /etc/lvmtab /etc/lvmtab.old
2)vgscan -av