Setting up Multiple Subnets

You can configure Platform to support multi-subnet architectures, in which each subnet uses its own PXE boot server. This is performed by installing Platform Server for the first subnet, then installing additional MaaS rack servers for each additional subnet. These additional rack servers register with the primary Platform Server, which is used for all normal operations.

Components

For a multi-subnet scenario, there are the following components and roles:

• 1 Region Server (your Primary server) which is a normal installation of the Platform Server.

  • This service is responsible for all configuration and management.

  • This server provides all services: DNS, HTTP Proxy, NTP, Syslog, etc.

  • The Platform Server may or may not manage its own storage subnet with storage nodes.

  • Required: It must be reachable by all rack servers and all storage nodes, regardless of which subnet they live on.

• 1+ Rack Servers to provide DHCP/PXE capabilities for the respective subnet.

  • The rack server does not provide any services except two:

    1. DHCP/PXE

    2. a service that provides cluster config to storage nodes when they boot

  • The storage config service does not contain any local config: it reads the config that is stored on the primary region server.

Versions — Verify packages with these 10.0 versions or higher are obtained:

• CaringoPlatformServer10.0*.ova

• CaringoPlatformRackServer10.0*.ova

• caringo-platform-10.0*.zip

Networking — Neither the primary region server nor any secondary rack servers participate in network routing.

Installing the Rack Server

Follow the method for your installation method:

Bare Metal

To install a rack server, you proceed as if for a primary region server with one change: When running the installplatform.sh script, you must include an extra flag (-r) that indicates the IP of the primary region server. The IP given must be reachable by both the rack server and its associated storage nodes.

1. Run the script with the -r flag:

   sudo ./installplatform.sh -i INTERFACE -r PRIMARY·SERVER·IP

2. Once this is completes, the CLI is installed on the rack server, and is automatically configured to point to the primary server.

3. From the rack server, use this command to see the new subnets that are managed by the rack server: 

   platform list subnets --all

OVA

1. Complete all necessary hostname and network configuration. 

2. Verify the OVA specific to the rack server is being used (which differs from the one for the primary region server), and import it. 
   Mac Addresses:
   Verify the OVA import process regenerates MAC addresses for any NICs if multiple subnets are used so no two rack servers have any MAC addresses in common.
   It is critical virtual environments generate new MAC addresses for the interfaces when the OVA (or equivalent) is imported. The installation stops if the installation detects another server with the same MAC address for any interface.

3. Run the following script:

   sudo /opt/reinitrack.sh -i PRIMARY·SERVER·IP

4. Use the list subnets command to view the new subnet that is managed by the current rack server in the list:

   platform list subnets --all

Setting up the Rack Server

After installation on a rack server, complete its setup:

1. Use the CLI to configure the new rack server to manage the new subnet. This is similar to the bootstrap command: 

   platform add dhcp -t SUBNET -a STARTING·DYNAMIC·IP -z ENDING·DYNAMIC·IP

2. Unless you have a non-standard DNS setup, point DNS back to the Primary server: 

   platform add dns -t SUBNET·FROM·ABOVE --dns PRIMARY·SERVER·IP

3. The Gateway for this secondary subnet is probably also provided by something else on the subnet. If so, set the Gateway for this subnet to point to that service: 

   platform add gateway -t SUBNET·FROM·ABOVE --gateway GATEWAY·IP