Product Documentation

Creating a Cluster

Mar 20, 2015

You can install multiple instances of the App Controller virtual machine (VM) to create a cluster. One App Controller VM acts as the cluster head. This App Controller is considered to be the host and, as such, hosts the database for all of the VMs in the cluster.

Note: The App Controller cluster head requires 4 VCPUs.

All other App Controller VMs in the cluster are called service nodes. Each service node has a local database that is used by the service node only. Updating user information from the service node to the cluster head requires writing to the database. A service node connects to the database on the cluster head by using a secure channel.

App Controller VMs deployed as service nodes obtain their configuration from the App Controller that acts as the cluster head. Citrix recommends deploying two App Controller VMs in a high availability pair. With high availability, one of the paired VMs serves as the primary node, acting as the cluster head, while the other is the secondary node, monitoring the primary. If the primary fails, the secondary assumes the role of the cluster head.

You use the command-line console to configure App Controller clustering. You can create a cluster, join an App Controller VM to a cluster, and remove a VM from the cluster.

When you add an App Controller service node to the cluster and then log on by using the management console, only the Dashboard and the home page appear.

Configuring Load Balancing for an App Controller Cluster

As shown in the following figure, App Controller works with NetScaler to provide load balancing to all of the service nodes in the cluster. All VMs run behind a load balancer that is responsible for terminating SSL connections from Citrix Receiver. You install certificates for App Controller on the load balancer.

Figure 1. Deploying App Controller with NetScaler in a Cluster

Citrix recommends that you configure the following three virtual servers on NetScaler for load balancing:

  • One virtual server as a Content Switching load balancer
  • One virtual server for rule-based load balancing
  • One virtual server for custom serverID persistence load balancing

You configure the Content Switching load balancer to route requests with the session ID in the URL to go through the custom serverID load balancer. All other requests go through the rule-based load balancer. You need to configure cookie (LB1) and serverID (LB2) persistency policies on the virtual servers. If the session ID parameter is not in the request URL, the request is sent to LB1 for cookie persistency. If the session ID is present, the request is sent to LB for serverID persistency.

The load balancer needs to query the cluster node to determine if the cluster is running, to obtain information about the load, or to perform health checks by sending user requests to App Controller VMs that are working correctly. Health checks provide greater application availability by ensuring user requests are directed only to correctly behaving servers.

When user connections go through the load balancer to an App Controller service node, the cluster uses HTTP connections. You can use HTTPS if you want additional security.

Note: App Controller clustering uses TCP port 9737

For more information about load balancing, see Load Balancing in the NetScaler documentation.

Creating a Cluster

When you create a cluster in the command-line console, the cluster node identifier, the current App Controller role, and a prompt to enter the shared key appear. You create the shared key on the cluster head and then use the same shared key for each VM in the cluster.

You must configure the host name on the cluster head. This information is replicated to the service nodes automatically as part of configuration sharing. The host name appears in the request URLs to be distributed in the cluster. Each service node in the cluster must also have the same shared key to establish secure tunnels. If the shared key on the cluster head and service node do not match, the two VMs cannot communicate.

The shared key must be eight characters with at least one uppercase letter, one lowercase letter, one numeric, and one special character that includes one of the following symbols: [!#$&].

After you enter the shared key, you restart App Controller. After App Controller restarts and you log on again to the command-line console, a message appears that states that the App Controller VM is now the cluster head.

If you have two App Controller VMs in a high availability pair, use the virtual IP address when you configure the service node. Both VMs function as a cluster head. For example, if the primary VM in the high availability pair fails for any reason, the secondary VM takes over and serves user requests.

Adding Service Nodes to the Cluster

When you join an App Controller VM to a cluster, you provide the IP address and shared key of the cluster head. After you enter this information, you restart App Controller. When App Controller restarts, you can use the command-line console to show the following cluster information:

  • Cluster node ID
  • Current role as the service node
  • Cluster head IP address
  • Shared key

Adding and Removing App Controller Nodes in a Cluster

To join App Controller virtual machines (VMs) together in a cluster, you first designate one VM as the cluster head. To do so, you use the command-line console to create the cluster. Then, you can add other VMs to the cluster. These VMs are designated as service nodes. You can log on to the command-line console in one of the following ways:

  • XenCenter Console tab.
  • vSphere Console tab.
  • Hyper-V Powershell.
  • Secure Shell (SSH) connection, such as from PuTTY.

To create the cluster head

  1. Log on to the App Controller command-line console.
  2. In the main menu, type 2 and then press ENTER.
  3. Type 4 and then press ENTER.
  4. In Cluster Head IP Address, type the IP address of the App Controller VM that is acting as the cluster head.
  5. In Cluster Shared Key, enter the shared key for the cluster and then press ENTER.
  6. When prompted, type y to restart App Controller.

To join a node to a cluster

You can remove an App Controller VM from the cluster at any time.

  1. Log on to the App Controller command-line console.
  2. In the main menu, type 2 and then press ENTER.
  3. Type 5 and then press ENTER.
  4. In Do you want to leave the cluster, press y and then press ENTER.
  5. When prompted, type y to restart App Controller.