Hi Jeremy,
> 1. keepalived manages the "master" server
> 2. "agent" on realserver connects to keepalived master
> 3. keepalived marshals agent into LVS grouping appropriately
> 4. keepalived adds agent into healthcheck subsystem
> 5. keepalived emails admins to announce new server
Note that there's a lot of overlap between keepalived and feedbackd here
- feedbackd does the first four of these points already.
However, feedbackd has no provision for backup directors - the agent can
only communicate with one master at a time. One idea here is to adapt
NECP to a multicast protocol to remove the need for separate
connections.
hey hey, I agree... Multicast rulez for scalability... I am interrested
into integrating feedback protocol into keepalived, but I don't know
currently if I will produce this code for OpenSource.
Best regards,
Alexandre
|