Cluster re-configuration

[kholikt]

I am trying to compile a procedure as I need to migrate my MC/SG machine SAN using different switch so all disk hardware path is going to change, and this I believe will affect the lock disk.

Pre-migration
1. cmhaltcl
2. vi /etc/rc.config.d/cmcluster and set autostart_cmcld to 0
3. cmquerycl -k -v -C /etc/cmcluster/my_cluster.config -n nodeA -n nodeB
4. Backup all config in /etc/cmcluster directory
# export out all MC/SG VG
5. vgexport -m -p -f -s -v

Post-migration
From the primary node
1. mkdir /dev/
2. mknod /dev//group c 64 0x000000
3. vgimport -v -m -p -f -s -v
4. vgchange -a y
5. vgchange -a n
6. Edit cluster configuration file as lock disk path has chanaged. FIRST_CLUSTER_LOCK_PV
7. cmcheckconf -k -v -C /etc/cmcluster/my_cluster.config
8. cmapplyconf -k -v -C /etc/cmcluster/my_cluster.config
9. vgcfgbackup
10. vgexport -m -p -f -s -v
11. rcp all map file to secondary node

From the secondary node
1. mkdir
2. mknod /dev//group c 64 0x000000
3. vgimport -v -m -p -f -s -v
4. vgchange -a y
5. vgchange -a n
6. verify the cluster lock disk configuration

Do I miss anything?

[Rajeev Shukla]

Your steps look good except for vgexport and vgimport.
If you use -p option with vgexport it wont remove the vg from /etc/lvmtab file and you'll have problems importing it.
So remove -p option and rest is all good.

[kholikt]

here is improve version.

1. Import all VG using the map files export in the pre-migration step.
# mkdir /dev/
# mknod /dev//group c 64 0x0X0000
# vgimport â v â m -p â f -s â v
# vgchange â a y

2. Deactivate the volume group on the primary node & copy LVM info to nodeB:
# ll /dev/vg*/group |awk '{print $6, $10}' > /tmp/gp_min.lst
# rcp /tmp/gp_min.lst nodeB: /tmp/gp_min.lst
# vgchange -a n
# vgexport -m -p -f -s -v
# rcp /tmp/ nodeB:/tmp/
# rcp /tmp/< pv_paths > nodeB:/tmp/< pv_paths >

3. Ensure the cluster lock device should have default pv timeout value on both Node

4. Modify/edit the following parameters in the cluster configuration file as lock disk path has changed:
FIRST_CLUSTER_LOCK_PV

5. Verify the cluster configuration:
# cmcheckconf -k -v -C /etc/cmcluster/emc_cluster.config

6. Activate the cluster lock volume group (if used), and apply the cluster configuration:
# vgchange -a y /dev/ (if cluster lock is used)
# cmapplyconf -k -v -C /etc/cmcluster/emc_cluster.config
If a cluster lock is used, back it up:
# vgcfgbackup /dev/

7. Deactivate cluster lock volume group:
# vgchange -a n /dev/

8. Start the cluster, check status, test node shutdown functionality, and halt the cluster:
# cmruncl -v
# cmviewcl -v
# cmhaltnode -v nodeA
# cmviewcl -v
# cmrunnode -v nodeA
# cmviewcl -v
# cmhaltcl -v â f


Secondary Node
1. Import the vg
# mkdir /dev/
# mknod /dev//group c 64 0x0X0000 (base on /tmp/gp_min.lst)
# vgimport -v -m /tmp/ -p -f /tmp/ -s -v
# vgchange -a y
# vgchange -a n /dev/vgdata1

2. Ensure the cluster lock device should have default pv timeout value on both Node

3. Modify/edit the following parameters in the cluster configuration file as lock disk path has changed:
Cluster lock device (or Quorum server) for each node

4. Verify the cluster configuration:
# cmcheckconf -k -v -C /etc/cmcluster/emc_cluster.config

5. Activate the cluster lock volume group (if used), and apply the cluster configuration:
# vgchange -a y /dev/ (if cluster lock is used)
# cmapplyconf -k -v -C /etc/cmcluster/emc_cluster.config

6. If a cluster lock is used, back it up:
# vgcfgbackup /dev/

7. Deactivate cluster lock volume group:
# vgchange -a n /dev/

8. Start the cluster, check status, test node shutdown functionality, and halt the cluster:
# cmruncl -v
# cmviewcl -v
# cmhaltnode -v nodeA
# cmviewcl -v
# cmrunnode -v nodeA
# cmviewcl -v
# cmhaltcl -v

[Stephen Doud]

Step 4 won't work - the cluster VGs are marked for exclusive activation, so a '-a y' conflicts with exclusive activation. You also cannot use '-a e' until the cluster daemons are running.

It may be necessary to add the following steps:
Step 6.5
# vgchange -c n
# vgchange -a y

Step 8.5
# vgchange -a n

[Geoff Wild]

Looks good except why do step 6 and 7?

6. Activate the cluster lock volume group (if used), and apply the cluster configuration:
# vgchange -a y /dev/ (if cluster lock is used)
# cmapplyconf -k -v -C /etc/cmcluster/emc_cluster.config
If a cluster lock is used, back it up:
# vgcfgbackup /dev/

7. Deactivate cluster lock volume group:
# vgchange -a n /dev/


You don't need a separate vg for the cluster lock....your cmquerycl will choose a device for you...

BTW - I'm about to do the same thing this weekend on one of my clusters...

Rgds...Geoff

[Calandrello]

Friend the Lock record moved? if yes you tera that to modify this field they tambem in the archive .ascii

[kholikt]

Hi,

As I mentioned earlier that all device path will change so I think I need to modify the lock PV configuration.

[kholikt]

Hi,

In my cluster configuration I have only the following line

FIRST_CLUSTER_LOCK_VG /dev/vgAA1

There is no
FIRST_CLUSTER_LOCK_PV so I guess I do not nee to do the following steps in both nodes right

4. Ensure the cluster lock device should have default pv timeout value on both Node

5. Modify/edit the following parameters in the cluster configuration file as lock disk path has changed:
FIRST_CLUSTER_LOCK_PV

6. Verify the cluster configuration:
# cmcheckconf -k -v -C /etc/cmcluster/emc_cluster.config

7. Activate the cluster lock volume group (if used), and apply the cluster configuration:
# vgchange -a y /dev/ (if cluster lock is used)
# cmapplyconf -k -v -C /etc/cmcluster/emc_cluster.config
If a cluster lock is used, back it up:
# vgcfgbackup /dev/

8. Deactivate cluster lock volume group:
# vgchange -a n /dev/

[kholikt]

sorry I have the FIRST_CLUSTER_LOCK_PV parameter in the config file I miss the line

[kholikt]

Here we go I have re-arranged the step, I think it should be correct now. Any comment are welcome. The scenario involve change the FIRST_CLUSTER_LOCK_PV. I am not too sure whether I need to cmapplyconf and cmcheckconf in one node only or both nodes

Primary Node
1. Import all VG using the map files export in the pre-migration step.
# mkdir /dev/
# mknod /dev//group c 64 0x0N0000
# vgimport â v â m -p â f -s â v

2. Activate MC/SG VG.
2.1 If cluster daemon is running use the following command:
# vgchange â a y

2.2 If cluster daemon is not running use the following command:
# vgchange â c n
# vgchange â a y

3. Deactivate the volume group on the primary node & copy LVM info to nodeB:
# ll /dev/vg*/group |awk '{print $6, $10}' > /tmp/gp_min.lst
# rcp /tmp/gp_min.lst nodeB: /tmp/gp_min.lst
# vgchange -a n

If you have done step 2.2, you have to convert the VG back to cluster mode.
# vgchange â c y

# vgexport -m -p -f -s -v
# rcp /tmp/ nodeB:/tmp/
# rcp /tmp/ nodeB:/tmp/< pv_paths >

Secondary Node
4. Import the vg using the map files copy from primary node
# mkdir /dev/
# mknod /dev//group c 64 0x0N0000
# vgimport â v â m -p â f -s â v

Primary Node and Secondary Noe
5. Ensure the cluster lock device should have default pv timeout value on both Node.

Primary Node
6. Locate and modify the following parameters in the cluster configuration file as lock disk path has changed:
FIRST_CLUSTER_LOCK_PV

7. Verify the cluster configuration:
# cmcheckconf -k -v -C /etc/cmcluster/emc_cluster.config

8. Activate the cluster lock volume group (if used), and apply the cluster configuration:
# vgchange â c n /dev/
# vgchange -a y /dev/
# cmapplyconf -k -v -C /etc/cmcluster/emc_cluster.config
# vgcfgbackup /dev/

9. Deactivate cluster lock volume group:
# vgchange -a n /dev/
# vgchange â c y /dev/

Primary Node & Secondary Node

10. Edit /etc/rc.config.d/cmcluster and set autostart_cmcld to 0

Primary Node

11. Start the cluster, check status, test node shutdown functionality, and halt the cluster:
# cmruncl -v
# cmviewcl â v

[Dinesh MIshra]

I thinks all the above steps are right.
only thing is u should activate the lock volume in exclusive mode(if required).

it will work.