Having a multi-node cluster of NSX Controllers helps ensure that at least one NSX Controller is always available.
Install three NSX Controller appliances.
Make sure the NSX Controller nodes have joined the management plane. See Join NSX Controllers with the NSX Manager.
Initialize the control cluster to create a control cluster master.
In the join control-cluster command, you must use an IP address, not a domain name.
If you are using vCenter and you are deploying NSX-T components to the same cluster, make sure to configure DRS anti-affinity rules. Anti-affinity rules prevent DRS from migrating more than one node to a single host.
- Open an SSH session for each of your NSX Controller appliances.
For example, NSX-Controller1, NSX-Controller2, and NSX-Controller3. In this example, NSX-Controller1 has already initialized the control cluster and is the control cluster master.
- On the non-master NSX Controllers, run the set control-cluster security-model command with a shared secret password. The shared-secret password entered for NSX-Controller2 and NSX-Controller3 must match the shared-secret password entered on NSX-Controller1.
NSX-Controller2> set control-cluster security-model shared-secret secret <NSX-Controller1’s-shared-secret-password> Security secret successfully set on the node.
NSX-Controller3> set control-cluster security-model shared-secret secret <NSX-Controller1’s-shared-secret-password> Security secret successfully set on the node.
- On the non-master NSX Controllers, run the get control-cluster certificate thumbprint command.
The command output is a string of numbers that is unique to each NSX Controller.
NSX-Controller2> get control-cluster certificate thumbprint ...
NSX-Controller3> get control-cluster certificate thumbprint ...
- On the master NSX Controller, run the join control-cluster command.
Provide the following information:
IP address with an optional port number of the non-master NSX Controllers (NSX-Controller2 and NSX-Controller3 in the example)
Certificate thumbprint of the non-master NSX Controllers
Do not run the join commands on multiple controllers in parallel. Make sure the each join is complete before joining another controller.
NSX-Controller1> join control-cluster <NSX-Controller2-IP> thumbprint <nsx-controller2's-thumbprint> Node 192.168.210.48 has successfully joined the control cluster. Please run 'activate control-cluster' command on the new node.
Make sure that NSX-Controller2 has joined the cluster by running the get control-cluster status command.
NSX-Controller1> join control-cluster <NSX-Controller3-IP> thumbprint <nsx-controller3's-thumbprint> Node 192.168.210.49 has successfully joined the control cluster. Please run 'activate control-cluster' command on the new node.
Make sure that NSX-Controller3 has joined the cluster by running the get control-cluster status command.
- On the two NSX Controller nodes that have joined the control cluster master, run the activate control-cluster command.
Do not run the activate commands on multiple NSX Controllers in parallel. Make sure each activation is complete before activating another controller.
NSX-Controller2> activate control-cluster Control cluster activation successful.
On NSX-Controller2, run the get control-cluster status verbose command, and make sure that the Zookeeper Server IP is reachable, ok.
NSX-Controller3> activate control-cluster Control cluster activation successful.
On NSX-Controller3, run the get control-cluster status verbose command, and make sure that the Zookeeper Server IP is reachable, ok.
- Verify the result by running the get control-cluster status command.
NSX-Controller1> get control-cluster status uuid: db4aa77a-4397-4d65-ad33-9fde79ac3c5c is master: true in majority: true uuid address status 0cfe232e-6c28-4fea-8aa4-b3518baef00d 192.168.210.47 active bd257108-b94e-4e6d-8b19-7fa6c012961d 192.168.210.48 active 538be554-1240-40e4-8e94-1497e963a2aa 192.168.210.49 active
The first UUID listed is for the control cluster as a whole. Each NSX Controller node has a UUID as well.
If you try to join a controller to a cluster and the command set control-cluster security-model or join control-cluster fails, the cluster configuration files might be in an inconsistent state.
To resolve the issue, perform the following steps:
On the NSX Controller that you try to join to the cluster, run the command deactivate control-cluster.
On the master controller, if the command get control-cluster status or get control-cluster status verbose displays information about the failed controller, run the command detach control-cluster <IP address of failed controller>.
What to do next
Deploy the NSX Edge. See NSX Edge Installation.