You can use NetCache configuration to specify one or more NetCache takeover pairs. Then, if one NetCache appliance in the pair goes down, the other takes over servicing HTTP, FTP, Gopher, MMS, and RTSP requests for the other appliance. This feature does not take over servicing NNTP requests.
For MMS and RTSP requests, existing streaming connections are lost when a NetCache appliance in the takeover pair fails. However, the takeover partner handles new streaming media requests for the appliance that failed.
The following illustration shows a NetCache appliance takeover pair.
On each appliance in the takeover pair, the other appliance is specified as its takeover partner. Every 10 seconds, each appliance then communicates with its partner by using a communication "heartbeat," similar to a ping, to determine whether its partner is alive. If a NetCache appliance does not respond to its partner's heartbeat query, the partner takes over and starts servicing all requests for the IP addresses and IP address aliases of its partner.
This takeover process enables clients of the down partner to continue to use the same IP address to contact a cache. Therefore, to most users and other NetCache appliances, it appears as if no failure occurred.
Each appliance in the takeover pair is equal from a failover perspective. There is no concept of a primary NetCache appliance and a secondary NetCache appliance, as there is with using the automatic proxy configuration file for Web caching.
If you are deploying transparent proxying with an L4 or L7 switch or a WCCP router, you would not use this failover mechanism because the switch or router provides better failover.
If you are deploying transparent proxying with a policy-based router, using this failover feature in combination with transparent proxying is a good strategy because currently available routers that use policy routing cannot provide failover.
This feature is supported for Ethernet and FDDI. It does not work with ATM.
Choose the simplest and most reliable connection possible between the two NetCache appliances. Network Appliance recommends that you use a crossover cable to connect the interfaces on the appliances that are the dedicated "takeover interfaces."
If you use another device, such as a switch, between the appliances, the risk exists that the device will fail, making this failover mechanism unavailable. If the connection between the two NetCache appliances fails, each NetCache appliance appears to the other to have failed and both appliances attempt to take over for each other. The result on your production networks is the same as if you had two appliances on the same network with the same IP address. Example of adding network fault-tolerance describes how you can add network fault-tolerance to your NetCache takeover scheme.
Although a NetCache appliance takeover pair provides higher system availability in case of a system failure, it does not ensure higher network fault-tolerance. If a network-related problem occurs, a NetCache appliance in a takeover pair might still become inaccessible.
The following illustration shows a deployment in which a single mode virtual interface (VIF) is combined with NetCache takeover to achieve a higher fault tolerance than can be achieved with NetCache takeover alone. The following discussion shows a single mode VIF used with a non-takeover interface Alternatively, a single mode VIF can be used with the takeover interface.
The previous illustration shows the failure zone for NetCache A (between Switch 1 and e0 on NetCache A, or the network between Switch 1 and NetCache A) that NetCache B cannot detect even though NetCache A might be inaccessible through the network.
To provide a fully redundant network architecture for your NetCache appliance takeover pair, you can set up a single-mode virtual interface on both takeover partners. Subsequently, if a failure occurs in the failure zone for NetCache A, the standby interface (e2) on NetCache A maintains connectivity through Switch 2. NetCache B can then take over caching services for NetCache A if NetCache A fails. Likewise, the standby interface on NetCache B would maintain connectivity to Switch 1 if a failure occurs in the failure zone for NetCache B. With this approach, if a network-related failure occurs, both NetCache appliances remain accessible through the network.
Note
In a single-mode virtual interface, all interfaces that constitute the virtual interface share an IP address. However, only one interface in the virtual interface is active. The other interfaces are on standby, ready to take over if the active interface fails.
The interface requirements for NetCache appliance takeover are as follows:
You must set up access controls the same way on both appliances in the designated pair. Otherwise, when one of the appliances fails and the other appliance takes over, the takeover partner does not apply the same controls. Likewise, if you are using user authentication, you must set up users the same way on both appliances. Alternatively, you can configure NetCache to use an LDAP or RADIUS authentication server or the NTLM authentication protocol. Otherwise, users who are not configured on the takeover partner are denied access if protocol authentication is turned on.
If you are using an automatic proxy configuration file for Web caching, you might have included entries to enable failover to another appliance. If so, the automatic proxy configuration file failover is not invoked unless both NetCache appliances in the pair go down. NetCache takeover takes precedence over other mechanisms.
Note
Do not enter the IP address of the dedicated takeover interface in an automatic proxy configuration file.
Copyright © 1994-2002, Network Appliance, Inc. All rights
reserved.
Part No. 210-00372
File: multipl5.htm
Last updated: 11/13/02 18:10:59