|
|
|
The two CNA systems should have nearly identical configurations, and these configurations should be kept synchronized.
If you change any command parameters on one, you should make a similar change on the other. This is especially important for any parameters that are used in the decision-making or load optimization process, such as (but not limited to) winner-set width, load threshold table values, or decision policies, for example.
Both CNA systems should establish BGP peerings with the edge routers, both should be configured to receive UTT and active-probe measurements from the same service provider links, and both should probe the same target list.
Both should poll the edge routers for usage (required for load optimization) and BGP status information at the same rates. Each device needs measurements and polled data in order to evaluate route conditions and optimize traffic for either performance or load.
Since data will not be shared between the two CNA systems, your edge routers must be polled by each CNA system independently of the other. This doubling of the polling burden on the routers is a necessary cost of using a second CNA system to achieve HA, since each device must be capable of optimizing network traffic on its own.
The only differences between the two CNA configurations should be the IP addresses assigned to the CNA module interfaces (module eth0 interfaces and USTAT VIPs, for example).
|
|
|