CSS 11503 Switch, Service Failure Detection and Fail-Over

[admin_2]

The configuration guides and command reference discribe the actions and timers available for "keepalive" support. I would assume "keepalive" is not the only mechanism that can change the state of a service to "dead", and that

repeated failures (e.g., fail to respond, checksum errors) of application packets, other than "keepalive" packets, could result in a Service being marked "dead".

Where can I find information that would describe the time it takes to discover a failures other then "keepalive" no response (e.g., time between retires, max retry count,...) and the configurable options on the CSS beyond those of the "keepalive" mechanism. In addition, after a Service is marked "dead" what happens to the failed request (packet) and ongoing HTTP (and TCP) session and other notions of stickiness? Are there options that support the Switch to automatically failover the request and continue the session on another available Service.

[didyap]

The normal way that the service is detecting if a service is alive, dying and finally dead is by the keepalives, if no response is received in the default values you get these status from the services.

Another way that I am aware of is that the CSS can send a SNMP trap in case a service is going down. But am not sure of the details.