MIRROR DISKS

[wayne_104]

HI

Can somone point me to a document that can help me out.

I have a 2 node cluster. We have 2 EVA's

I have presented the disks on the 2 eva's to the nodes.

I now want to mirror the disks on eva 1 and eva 2 to node one and then do the same on node2.so i have a mirroed disk with data residing on eva 1 and eva2.

If there is a failure on eva one the cluster will still see the data on node2.

If there is now a failure on say node1 the cluster will fail to node2 and see the data on eva2.

[Torsten.]

I would simply create equal LUNs on both arrays and use LVM to mirror the LVOLs.

In case of an EVA failure the server will continue with the other EVA, in case of a cluster switch the other node will activate the VG and run the application.

[wayne_104]

As i said . on the eva i have presented equal partiotions to both nodes.

on node one i have done the mirror and lvextended the vg. no problems when i switch to the second node does not like the vg.

[Torsten.]

>> second node does not like the vg

Details? Messages?

Did you vgimport the modified VG into the second node?

[likid0]

What you are saying is what they call extended cluster(old campus cluster in HP), check the docs site you have a lot of info, on how to get it working

[Stephen Doud]

You did not say what sort of problem the other node had when you had it attempt to access the VGs. That would be helpful to know.

As others have described, create a volume group that spans both EVAs, and mirror LVOLs on one EVA to the other EVA. Having done this, you may need to vgexport and reimport the VG on the other node, since the LVM commands to create the VG, lvol and mirrors were only performed on one node.


DETAIL:
The /etc/lvmtab file is the map that LVM uses to activate a volume group. Some of the content of the binary file can be viewed using 'strings /etc/lvmtab'. If you find a discrepancy in the number if disk special files (DSF) per VG across nodes, then you need to re-import the VG on the node that has fewer DSFs. The procedure is described in other threads.

Once /etc/lvmtab lists all EVA LUNs with each VG, a single EVA failure should be transparent when LVM switches to alternate links.

[wayne_104]

Got it working.

Halt the cluster
#cmhaltcl â f
#vi /etc/rc.config.d/cmcluster
AUTOSTART_CMCLD=0


# pvcreate -f /dev/rdisk/disk38
OUTPUT
Physical volume "/dev/rdisk/disk38" has been successfully created.

# vgchange -c n vg02
Output
Performed Configuration change.
Volume group "vg03" has been successfully changed.

# vgchange -a y vg03
Output
Activated volume group
Volume group "vg03" has been successfully changed.

# vgextend /dev/vg02 /dev/disk/disk23

OUTPUT
Volume group "/dev/vg02" has been successfully extended.
Volume Group configuration for /dev/vg02 has been saved in /etc/lvmconf/vg

#lvextend -m 1 /dev/vg02/lvol1 /dev/disk/disk23

OUTPUT
The newly allocated mirrors are now being synchronized. This operation will
take some time. Please wait ....


DEACTIVATE DISK
#vgchange -a n /dev/vgnn

# vgexport -v -p -s -m vg02.map /dev/vg02

# scp vg02.map node2:/

OUTPUT
The authenticity of host 'alsmt1 (10.192.100.19)' can't be established.
RSA key fingerprint is 60:5b:0d:95:fc:aa:60:f9:42:6a:1c:a1:75:a2:fe:f7.
Are you sure you want to continue connecting (yes/no)? yes
Warning: Permanently added 'alsmt1,10.192.100.19' (RSA) to the list of known hosts.
Password:
Vg02.map 100% 22 0.0KB/s 0

ON SECOND NODE

# mkdir /dev/vg02
# mknod /dev/vg02/group c 64 0x020000 NOTE THE NUMBER
# vgimport -v -s -m /vg02.map vg02


â > Agile Multipathing of HP-UX 11.31 is not used by default after import (HP-UX 11.31 Bug ?!). The volume group uses alternate LVM Paths.
Solution:
# vgchange -a n vg02
# vgchange -a y vg02

// Add agile Path
# vgextend /dev/vg02 /dev/disk/disk39
NOTE THIS DISK NUMBER MAY BE DIFFRENT ON THE SECOND NODE

// Remove Alternate Paths
# vgreduce vg02 /dev/dsk/c16t0d1 /dev/dsk/c10t0d1

OUTPUT

Device file path "/dev/dsk/c16t0d1" is an primary link.
Removing primary link and switching to an alternate link.
Device file path "/dev/dsk/c10t0d1" is an alternate path.
Volume group "vg02" has been successfully reduced.
Volume Group configuration for /dev/vg02 has been saved in /etc/lvmconf/vg02.conf

BACKUP VG
# vgchange -a r vg02
# vgcfgbackup /dev/vg02
Volume Group configuration for /dev/vg02 has been saved in /etc/lvmconf/vg02.conf

Deactivate mount

#vgchange -a n /dev/vgnn for ll vgâ s
#vgchange -c y /dev/vgnn for all vgâ s


Change in /etc/cmcluster/IALCH_oraDB/IALCH_oraDB.conf

#vi /etc/cmcluster/IALCH_oraDB/IALCH_oraDB.conf
vgchange_cmd "vgchange -a e -q n"

APPLY PACKAGE

#cmcheckconf -P $SGCONF/IALCH_oraDB/IALCH_oraDB.conf

If No errors the apply package.
#cmapplyconf -P $SGCONF/IALCH_oraDB/IALCH_oraDB.conf
OUTPUT
Validation for package IALCH_oraDB succeeded via /etc/cmcluster/scripts/mscripts/master_control_script.sh.

Modify the package configuration ([y]/n)? y
Completed the cluster update
#vi /etc/rc.config.d/cmcluster
AUTOSTART_CMCLD=1
Reboot systems

[Matti_Kurkela · Accepted solution]

# vgimport -v -s -m /vg02.map vg02

>> Agile Multipathing of HP-UX 11.31 is not used by default after import (HP-UX 11.31 Bug ?!). The volume group uses alternate LVM Paths.

Not a bug. It's just that 11.31 uses the legacy devices unless you explicitly request to use the new ones. I guess this is HP's way to maximize compatibility with legacy scripts and software.

You should have used the -N option:

# vgimport -N -v -s -m /vg02.map vg02

MK

[wayne_104]

see above

[wayne_104]

Thanks Matti

gave you 10 pointer for that