4.4. Servers
After you have determined what services will be running on the cluster, you can configure the servers that will be running those services. These servers are the cluster nodes that will process the service requests. Server configuration is divided into two parts:
Servers Configuration
Server Groups Configuration
4.4.1. Servers Configuration
The first thing you'll need to do is create the list of servers that will be used. Here's how:
From the tlclbconfig main menu, choose `Servers Configuration'.
Choose `Servers Configuration' again from the next menu. This will show a list of all the cluster nodes that are to be used in the cluster. If you do not have service nodes configured initially, you will not see a list in the window.
Each line in the list has the name of the node, which forwarding method it is using, and a flag indicating if it should be checked periodically using ping.
From this list, you can add a cluster node, edit an existing one, or remove a node.
To add a cluster node server, click `Add'.
There are three fields to fill in, plus a flag.
The `Server name' field is used to give the node a label, so it can be referred to later. It is best to use the hostname of the node.
Enter the IP address or fully qualified domain name of the cluster node on the line labeled `Full server name or IP'. As we noted above, the address must resolve forward and backwards, even if you enter it by IP address. It is probably best to specify the address by name.
Select the `Forward method'. Your choices are `direct', `tunnel', and `nat'. We'll cover each of these in more detail below.
If, for some reason, you don't want the ATM to check the node periodically to see if it is alive, uncheck the `ping to see if alive' box. You should normally leave this checked so that the ATM can remove the node from the cluster if it goes down.
When you have entered the information about the server node, click `OK'.
After you have added information about all the cluster nodes, click `Done'.
Don't forget that a server node can also act as a primary or backup ATM.
4.4.1.1. Forwarding Mechanisms
The three traffic forwarding mechanisms are direct forwarding, tunneling, and NAT. Each server will have one of these three forwarding methods. The choice of forwarding method is determined by several factors, including the location of the system, the operating system, and the amount of configuration you want to perform on the node.
The primary ATM actually uses a subclass of direct forwarding called "local", no matter which forwarding mechanism it was configured to use. This is because the incoming packets are delivered locally; they will not have to be actually forwarded anywhere if they are destined for the ATM system itself.
Except for NAT, cluster nodes must be configured so that they can be used in the cluster. The particulars depend upon which forwarding method is used and what operating system the node runs. For complete information, see Chapter 5.
4.4.1.2. Direct Forwarding
Direct forwarding is the default forwarding mechanism. It has the least overhead, but may require some modification of some settings on the cluster node. When an incoming packet is received from a client, the primary ATM forwards the packet to the cluster node. With direct forwarding, the packet is forwarded just like any other packet would be transferred.
When the cluster node server responds to the client request, it sends the response directly to the client. The return traffic does not need to go through the ATM.
Cluster nodes using direct forwarding must be located on the same LAN segment an subnet as the traffic manager. Direct forwarding should work with nodes running just about any operating system.
4.4.1.3. Tunneling
If a cluster node is on a different LAN than the ATM, you should use the tunnel forwarding method. The packet will be encapsulated and sent through a tunnel point-to-point connection between the ATM and the node. The tunneling method only works with cluster nodes running Linux or UNIX variants that support IP-IP tunneling.
The tunneling method incurs a small amount of overhead due to the encapsulation process. Like direct forwarding, responses are sent directly to the client and do not need to go through the ATM.
4.4.1.4. NAT
NAT stands for Network Address Translation. This method allows you to use systems as cluster nodes without having to do any special configuration on the nodes. This means that almost any system can be used as a cluster node when using NAT. However, extra attention must be given to configuring the ATMs.
NAT is the least efficient of the three forwarding mechanisms. There is a larger amount of overhead required to translate the client addresses. Also, return traffic must be routed through the ATM. This can become a serious bottleneck, because service response packets are often much larger than the request packets.
For more information on NAT forwarding, see Section 4.7.3.
4.4.2. Server Groups Configuration
Server groups provide a way to collect several server nodes into a group in order to add them to the cluster. This is also where you define the services that each server node will perform. Every server does not have to provide every service; you can selectively choose which servers run which services, and even configure some servers to take a larger share of the workload. Server groups are sometimes also referred to as server pools.
To set up a server group, follow these steps:
From the tlclbconfig main menu, choose `Servers Configuration'.
Choose `Server Groups Configuration' from the next menu. This will show a list of the server pools that have been defined. Initially, there will be no server groups, so you will have to create one.
Click `Add' to create a new server group.
This will pull up a new form, as shown below.
Skip down to the data entry fields -- we'll add servers to the pool after we define the server pool settings.
First give the server group a name, in the `Server pool name' box.
The `Check server' section has two parameters that you can set: the `Frequency' and the `Timeout'. Both are defined in seconds. The `Frequency' specifies how often the server nodes should be pinged to see if they are alive. The `Timeout' value tells how long to wait for a response before assuming that the server node is down. The `Timeout' must be shorter than the `Frequency'. You should accept the default values.
The `Check service' section is similar to the `Check server' section. It also has `Frequency' and `Timeout' settings. These settings tell the ATM how often to run the ASA, and how long to wait for a response before marking the service on the individual node as being down. Again, make sure the `Timeout' is shorter than the `Frequency'.
You should generally accept the default values for the server and service checks. You may want to fine tune them later to optimize the performance of the cluster. See Chapter 7 for more information on tuning the cluster.
Now that we have defined the parameters for the server group, we need to add some servers to the pool.
Go back to the top of the form and click `Add'.
Choose a server from the list and click `OK'. The list should contain all the server nodes that you configured in the `Servers Configuration' menu. Scroll through the list if you don't see the node you are looking for.
Another window will pop up, allowing you to select which services the node will run. This list will appear on the left. To the right are buttons to `Add', `Edit', and `Remove' services from this particular server.
Click `Add' to add a service for this server.
Choose a service from the list. The list scrolls if there are too many to fit in the window.
Set the `Weight' of this server when performing this service. Normally you will leave this at 1 for all servers. However, if you want some servers to receive a larger workload than others, you should give the more powerful servers a higher weight.
Click `OK' to add the service to the server.
Add any additional services for the server node you are editing, then click `OK'.
Click `Add' again to add other servers to the pool.
Click `OK' when you are finished configuring the server group.