Extreme Networks Px Series manual Configuring Server Groups

answer requests. Servers can share an IP address, but the port must be unique for each server.

•Max-connections represents the maximum number of concurrent connections this server can handle. After that number is reached, no more connections are sent to that particular server until some of the open ones have been closed (unless a persistence method is specified; see “Persistence Modes” on page 6-2). Most servers can handle fewer than 5,000 connections.

•Weight is used by weighted algorithms for load balancing. Use equal weights for all servers (or 1 for simplicity’s sake) with non-weighted algorithms. See Chapter 6.

If servers are configured at contiguous IP addresses, and have identical attributes, you can specify many identical servers at once using a range of IP addresses. The specified index is used for the first server and incremented for each configured server. The following example creates servers with indexes 3 through 10:

config server index 3 ipaddress 10.2.2.2 - 10.2.2.9 port 80

max-connections 4000 weight 1

To remove a server or range of servers from the system, use the following commands:

unconfig	server	index	<index>
unconfig	server index		<index> - <index2>
unconfig	server	ipaddress		<ipaddress>
unconfig	server	ipaddress		<ipaddress> - <ipaddress>

Configuring Server Groups

After all of the servers needed for a particular virtual service have been created, they must be organized into a server group. This group is used in the definition of the virtual service itself. The following command creates a server group:

config server-group name <string> policy [rr wrr lc wlc]

Each server group definition includes a unique name, used when configuring the server group or used elsewhere in the configuration. A load balancing policy is the method of choosing servers. See Chapter 6 for policy details.

You can optionally specify a server of last resort for the group. This is a server to which traffic is sent if all the servers in a server-group are down. It could be a server that simply replies to the client with a "SYSTEM DOWN" message, or a server that can service the request under emergency circumstances (perhaps a development machine or

5-2 Px Series Application Switch Installation and Configuration Guide