Replace Server on cluster

[Fernando Boza]

I have a cluster with HP-UXv1 with 2 nodes.
1- rp7420
1- rp7410

I want to replace the rp7410 with another rp7420.
What is the best process to do??
What considerations should we have?

[Vishu]

Hi Fernando,

Please tell how your current cluster is defined.

1) how many package are configured in the cluster.
2) is rp7410 the primary node for the package or the failover node for the package.

[Suraj K Sankari]

Hi,
>>I want to replace the rp7410 with another rp7420.
>>What is the best process to do??

The best option is take Ignite backup or rp7410 and restore into other box. After restoration do the necessary changes.


Suraj

[Raj D. · Accepted solution]

Fernando,

>>I have a cluster with HP-UXv1 with 2 nodes.
1- rp7420 -- say it is server: A
1- rp7410 -- say it is server: B

>>replace the rp7410 with another rp7420 -- say C.




- Build server C with same config as B , and same lan connection & subnet.
- Take ignite tape/network image of B and restore image on C.
- Move all pacakges of B server to A.
- Halt the node B.
- Remove the node B from cluster.ASCII file and package config files.
- Compile the pkg and ascii file:

#cmcheckconf -v -P pkg.conf
#cmapplyconf -v -P pkg.conf

#cmcheckconf -v -C cluster.ascii
#cmapplyconf -f -v -C cluster.ascii

That will make it a single node custer.


- Halt B system .

- Configure the ip of B system to C system (new target rp7420)
- Zone all the pkg vgs of B system , to new CC system, make the luns visible.
- create vgdir /group files/ copy map files / import the pkg vgs to this system.

- create the users that supposed to be on system C, and the pkg users.


- copy server A : /etc/cmcluster/pkg[*] directories to system C.

- Add the hostname of server C in cluster. ascii file. Make sure the lan parametrs are configured correctly as it was in B , or edit ascii file for C's lan parameters.

- compile cluster.ascii file , make sure quorum server or lock/disk mentioned properly.
#cmcheckconf -v -C cluster.ascii
#cmapplyconf -v -C cluster.ascii

- At this point the new rp7420 (system C) will be part if the 2 node cluster. A & C.


- Add system C's node name in pkg.conf file and compile the pkgs.
# cmcheckconf -v -P pkg.conf
# cmapplyconf -v -P pkg.conf

- check with cmviewcl.



- Bring back the packages to the new node C .
# halt its packages on A.
# start its packages on C.

- A good idea would be to check the FO_test of node A packages as well to this new node (C) if applicable.

-verify with cmviewcl and check vg/lv/fs's and application(s).

-Done.


Hth,
Raj.

[Raj D.]

Fernando,
Also further refrence :
http://docs.hp.com/en/B3936-90100/B3936-90100.pdf

Cheers,
Raj.

[Fernando Boza]

Hi Ranju

The rp7420 is the Node1 and rp7410 is Node2

Packages on Node1:
- Data Base Sybase
- Aplication "Monitran"

Packages on Node2:
- Hitachi ShadowImage (split replication of discs on Hitachi storage and then present that volumens on Node2 for backup)

[Fernando Boza]

Hi Suraj K Sankari

What are the necesary changes?

[Vishu]

okay, now i am assuming that your OS is installed on the new box.

You have to go this way.

1) Move all the packages from Node B to Node A.
2) Halt your node B.
3) Move your node B out of the cluster config by editing the cluster.ASCII and package.conf files and reapllying the new configuration with checkconf and cmapplyconf.

Now, your cluster is having only one node A and your other node B is out of the cluster.

4) Now, down the node B.
5) configure your node C(new server) with same configurations as your node B was having i.e. Ip address, VG creation, user creation, /etc/hosts, /etc/group, /etc/resolv.conf.
6) copy /etc/cmcluster/* from node A to node C.

Your server is now ready to add to the cluster
7) Edit the cluster.ASCII and package.conf to add node C in them. Make sure to take care of cluster lock disk configuration.
8) reapllying the new configuration with cmcheckconf and cmapplyconf.

Now you have two nodes (A and C) cluster again.

9) test the cluster by restarting the cluster and by switching packages from one node to other.

[Tor-Arne Nostdal]

just a small question...
why wouldn't you put the new server C as a third node ?
This way you could still let your primary/failover be intact or what...

[Fernando Boza]

FYI.
The package "ShadowImage" run only on Node2

[Steven E. Protter]

Shalom,

The hardware is very similar.

Therefore.

Get Ignite up to date on an Ignite server and the rp7410

make_net_recovery (make_net_recovery can be used if both systems have supported tape drives)

Move the new rp7420 into place. Make sure it has access to the heartbeat networks and all.

boot the 7420 from the Ignite server and conduct an install based on the rp7410 image.

Make changes to deal with hardware differences.

It should, if it has access to shared storage, jump right back in the cluster when done.

SEP