Each NSX appliance cluster – Global Manager, Local Manager or NSX Manager – must contain three VMs.
Those three VMs can be physically all deployed in the same data center or in different data centers, as long as latency between VMs in the cluster is below 10ms. In a 3-node Manager cluster, only one of the nodes can fail at any given time for the cluster to self-heal.
Single Site Requirements and Recommendations
The following recommendations apply to single site NSX deployments and cover cluster recommendations for Global Manager, Local Manager and NSX Manager appliances:
- Place your NSX appliances on different hosts (from anti-affinity rules in vCenter Server) to avoid a single host failure impacting multiple managers.
- Maximum latency between NSX appliances is 10ms.
- You can place NSX appliances in different vSphere clusters or in a common vSphere cluster.
- It is recommended that you place NSX appliances in different management subnets or a shared management subnet. When using vSphere HA it is recommended to use a shared management subnet so NSX appliances that are recovered by vSphere can preserve their IP address.
- It is recommended that you place NSX appliances on shared storage also. For vSphere HA, please review the requirements for that solution.
You can also use vSphere HA with NSX to provide recovery of a lost NSX appliance when the host where the NSX appliance is running fails.
Scenario example:
- A vSphere cluster in which all three NSX Managers are deployed.
- The vSphere cluster consists of four or more hosts:
- Host-01 with nsxmgr-01 deployed
- Host-02 with nsxmgr-02 deployed
- Host-03 with nsxmgr-03 deployed
- Host-04 with no NSX Manager deployed
- vSphere HA is configured to recover any lost NSX Manager (e.g., nsxmgr-01) from any host (e.g., Host-01) to Host-04.
