I'm in the midst of planning a heartbeat switch migration and will need to iron out a few kinks. My environment consists of multiple VCS cluster groups, running on Windows 2003/2008. Apparently one of the network switch connecting all the heartbeats in need of replacement(it is still online, it just needs to be replaced due to end of support).
My headache is on how to perform the migration without interrupting the clustered applications. Each of the cluster nodes are running on 2-3HB adapters, so initially I thought that replacing them one-by-one will work flawlessly. But when I tested it on a test environment, apparently after I removed and reconnect a HB connection from each pair of nodes, the nodes' state still remain as "Jeopardy" even after reconnection of HB cables.
Do you have any idea on what's the best method to migrate all HB connections to a different switch online?
What I've did during testing :
1. Freeze all service groups
2. Unplug connections for HB1 for both NodeA and NodeB from oldSwitch and connect them to newSwitch. Both NodeA and NodeB turns to "Jeopardy" at this stage and didn't return to normal state even after reconnection to new switch.
3. I didn't proceed to perform the same on HB2 connections for both node due to my concern with that "Jeopardy" state, afraid that if I were to proceed, it will make two mini-clusters instead.
4. What I did to return both node from Jeopardy state is to run Cluster Configuration Wizard on NodeA and reconfigure LLT connection to reset communication between the HB adapters.
Appreciate any advise. Thanks in advance.