ServiceGuard Configuration Help...

[Lee Harris_5]

Hello,

It's too hot today, and my brain hurts. Can someone just give me a few pointers here and just sanity check my plan please?

I need to take a node out of a two node SG cluster, and replace it with a new server. The new server is currently out of the cluster and has been setup as follows:

HP-UX 11.11 install and patching.
Oracle 9.2.4 install.
Local filesystems configured.
SG cluster config ascii files and control scripts etc all in place.

Let's say the current cluster consists of NODE1 and NODE2. NODE1 is running the packages. The new box I will call NODE3 for easiness (is that a word?)

My plan is as follows:

vgexport -p -v -s -m mapfile /dev/vgxx the shared volume groups on NODE1. Copy the files to NODE3.

cmhaltnode on NODE2

Shutdown NODE2, disconnect it from the LAN and SAN.

Unrack NODE2

Rack NODE3 in NODE2's place.

Bring up NODE3. Configure it with NODE2's hostname and IP addresses.

Connect NODE3 to the LAN and SAN.

vgimport -s -v -m mapfile /dev/vgxx from the mapfiles taken off NODE1.

cmrunnode NODE3.

How does that sound. It sounds a little bit TOO simple to me, I'm sure I've overlooked something.

Thanks - Lee

The cluster at the minute has node1 and node2. Node1 is running the packages. I plan to cmhaltnode node2, sh

[melvyn burnard · Accepted solution]

You are missing the fact you will need to reapply the binary with cmapplyconf.
The new server will (I assume) have it's own LAN cards, and also device addresses for things like the cluster lock disc.
As the binary contains such vital information, you will need to update it with the MAC addresses of the new cards etc.

You also need to make sure that the SG version and patch level is the same as the current cluster

[freddy_21]

Make sure about patches for SG, SG version and ur aplication side.

the step are:
1. create map file at server 1
2. stop cluster
3. shutdown server 2 and replace with server 3
4. import map file
5. you must edit cluster ascii file cmclconfig.ascii for network interface and cluster lock disk. why you must do that
- Your hardware path for lan card maybe not same with the oldest.
- cluster lock disk maybe different with the oldest

6. check/apply cluster configuration with cmcheckconf/cmapplyconf command.

7. check and apply package configuration.

The changes only at cluster configuration. when you apply cluster configuration all package must be down.

[Lee Harris_5]

Thanks for the advice so far. Is there anything special I need to do when removing NODE2?

This is my plan based upon your input...

vgexports on node1.
cmhaltnode on node2
shutdown -hy now on node2
Bring up node3 and configure it with node2's IP / hostname.
Edit cmclconfig.ascii (or whatever)
cmhaltcl
cmcheckconf -C cmclconfig.ascii
cmapplyconf -C cmclconfig.ascii
cmruncl

Do I need to do anything else special when effectively removing NODE2 or am I right in thinking that the re-application of the cluster config is sufficient?

Thanks

[freddy_21]

testing manually for your aplication. startup manual filesystem at node 3 and run startup script at control file.

this step to minimize problem at your cluster.


Good Luck
Freddy