How to define the subnet of inter-comm link?

Commercial Modules
1

i am going to install HA in an environment that does not allow a physical cable for heartbeat link.
What i see now during the setup is a random class B subnet. With that how can i define a subnet that is part of the same CIDR?

2

You can’t. HA requires the heartbeat to be on a different subnet.

I think you’re saying ‘my two HA nodes are on different subnets’, and that’s not going to work either.

Both machines must be on the same network segment.

Edit: I realise I’m using similar words that have totally different meanings here, so just to clarify:

A network segment is a single VLAN or network, or switch, that has a router on it. All the machines on that segment can talk to each other without going through a router.

A network subnet is an address range. 192.168.1.0/24 is a subnet, 10.0.0.0/8 is a subnet, and 172.16.0.0/16 is a subnet.

Multiple subnets can be on the same segment.

3

both the HA nodes will be on the same network segment with a different set of subnet. Here is what we intend to use:

HA nodes: 172.31.2.0/24
Heartbeat: 172.31.4.0/24

Right now it’s greyed out. But if we can define it, it will make it easier for us.

4

Please do not try to change the management link. It’s designed to be read only and unmanageable. The cluster manages it itself, and uses it to talk between the nodes on an explicitly private network.

There is no need for you to care about it, or even know it exists 8) (Unless you’re already USING those IP addresses, in which case you’re not going to have much fun, but the cluster tested the viability of the addresses when it created itself, and they weren’t in use then)

–Rob

5

i wish to do that too. since the environment does not allow a direct cable, i am using one ethernet port for all the communications.

the switch will provide a VLAN: 172.31.0.0/16. For my case, the cluster create an IP that is outside of this VLAN, an example is 172.25.xx.xx/24.

After the setup is complete, node A will start to fail as it could not contact the MySQL which is in 172.25.xx.11/24.

Since i can have other VLAN in the /16, i clone node A and assign another subnet that allow communication. However, that lead to another issue where node B detect that the cluster was not setup properly.

I would like to inch forward. So you mentioned it’s not going to be much fun. Will it be easier if i can change the subnet before the the syncing/migration/drdb assign the appropriate IP addresses in the configs?

6

The Switch provides a segment. Not a subnet. This is where you’re confused. The configuration, as it is, will work perfectly, if the interfaces are on the same segment. Please disregard the IP addresses of the internal network.

Edit: If the switch is filtering based on IP addresses, that’s 100% not going to work, at all, ever, even if you change the IP addresses to match, as other multicast and broadcast traffic is sent on that segment. You must have the machines on the same segment. That is the only requirement.