Solved

Two clusters sharing the same subnet for network segmentation ?

  • 3 February 2020
  • 3 replies
  • 2446 views

Hello,


I’m trying to configure two AHV clusters (3 nodes) for replication.
Both environments are running AOS 5.11.2 on HPE DX360 Gen10.
 

The network environment is spread between two stacks: one stack dedicated to the storage with 10Gb interfaces and one stack for the LAN with 1Gb interfaces. There is no connection between these two networks.
After the deployment of the nodes and the configuration of the clusters, I start to configure the Backplane Network with the following parameter:
Subnet: 172.16.250.0
Netmask: 255.255.255.0
VLAN: 202
Host node: br1

The backplane network of the first cluster works without any problem.
When I do the same (with the same) for the second cluster, it fails. It says that the IPs addresses are already in use. 

I would like to specify a VLAN/subnet where both clusters have their backplane IPs so I can use the backplane network for the replication between both clusters.

 

I hope this is clear.

 

Kind regards,

 

Fred 

icon

Best answer by bbbburns 7 February 2020, 16:07

View original

This topic has been closed for comments

3 replies

Hello,

I believe this issue is because both cluster’s backplane network are in the same subnet.

For new backplane networks, you must specify a non-routable subnet. The interfaces on the backplane network are automatically assigned IP addresses from this subnet, so reserve the entire subnet for the backplane network alone.

Also, If you plan to specify a VLAN for the backplane network, make sure that the VLAN is configured on the physical switch ports to which the nodes are connected.

Reference: https://portal.nutanix.com/#/page/docs/details?targetId=Nutanix-Security-Guide-v5_16:wc-network-segmentation-wc-c.html

Userlevel 1
Badge +1

@fredzone.net 

  1. Backplane networks are non-routable subnet and cannot be used for replication, backplane networks are purely for CVMs / Hosts and for Intra cluster communication only and not for intercluster.
  2. In your case, you got to use different subnet/VLAN for other cluster.

And the failure you are observing is as per design, more likely.

Userlevel 3
Badge +14

@fredzone.net I would recommend changing the subnet mask so you’re able to use the same VLAN for multiple clusters if you desire. This allows you to dedicate a subnet for the backplane network of each cluster, and avoid overlapping the backplane IP addresses.

For example:

Cluster 1 Backplane:

Subnet: 172.16.250.0
Netmask: 255.255.255.128
VLAN: 202

Usable IPs: 172.16.250.1 - 172.16.250.127

 

Cluster 2 Backplane:

Subnet: 172.16.250.129
Netmask: 255.255.255.128
VLAN: 202

Usable IPs: 172.16.250.129 - 172.16.250.254

 

That gives you the ability to add 127 nodes in every cluster - however - that may not be practical so you could subnet that even further with a smaller netmask if you require.