Problem(Abstract)
=================
On occasion, the entire cluster IP topology has to be reconfigured to accomadate: a subnet change, a netmask change, or simply all IP addresses used are changing.
Resolving the problem
======================

Procedure to reconfigure cluster topology:

1) Make a cluster snapshot and use the
/usr/es/sbin/cluster/snapshots/*.info as a reference.

2) stop cluster services on all nodes of the cluster: smitty clstop
(resources offline).

3) Use smitty chinet to change the IP address (or netmask if needed)
on the AIX en# network interfaces.

4) cp /etc/hosts /etc/hosts.bak
vi /etc/hosts and change the names and IP addresses to match the
changes you are made to the interfaces.

cp /etc/resolv.conf /etc/resolv.conf.bak
vi /etc/resolv.conf and update the domain name and nameserver
address(es) if they are changing.

5) cd /usr/es/sbin/cluster/etc
cp rhosts rhosts.bak
vi rhosts and change all addresses to match the addresses now
in /etc/hosts and on the interfaces.

6) You must also change the Communication Path to Node (similar to pg.
52 (pdf 60) in the 6.1 Admin. Guide):
smitty hacmp

In SMIT, select Extended Configuration > Extended Topology
Configuration > Configure HACMP Nodes > Change/Show a Node to the
HACMP Cluster and press Enter.

Press F4 on Communication Path to Node and select the IP address and
name from the pick list which matches the HA-nodename.

7) rm /usr/es/sbin/cluster/etc/config/clip_config
This will cause no harm and the file will be recreated in the next
step.

8) Removing an HACMP Network is detailed in the
Administration Guide on pg 296 (PDF 304).

Remove the HA-defined network which contains the IP addresses with
the incorrect netmask.

Do not attempt to synchronize yet.

9) See pg 51: Discovering HACMP-Related Information
In SMIT, select Extended Configuration > Discover HACMP-related
Information from Configured Nodes and press Enter.

Note: all the interfaces defined to HA have to be up and pingable
and the clcomdES subsystem (clcomd daemon) have to be running on both
nodes for this to work (restart clcomdES after changing the rhosts
file).

10) See pg. 53 (pdf 61), Configuring IP-Based Networks
Add the “net_ether_0#” network.
(this is to add the network back that we removed in step 6)

11) See the section “Configuring communication interfaces/devices to
HACMP” on pg. 57, sub-section: Configuring Discovered Communication
Interfaces to HACMP and add the appropriate inteface(s) to the
appropriate network.

12) See pg. 144 (pdf 152), Verifying the Cluster Using the Extended
Configuration Path. Use the default parameters except:
Automatically correct errors found during verification? [Yes]

Advertisements