RFD 95 Discussion about rfd HOT 1 OPEN

@jclulow I have a few questions on how some of this might work. Specifically, is this supervisor intended to take over all the supervisor duties that SMF would normally take? If this crashes what is supervising it (SMF)? In the event that it crashes does it take down any haproxy children that it has and if so how is that configured? Can you also share how the contract as a whole would reach to fatal signals in this regard? One of the things I want to make sure we avoid is the classic ctrun noorphan case which creates a different contract for its children, thus causing things like svcadm restart -d to be mostly useless.

While today adding a NIC causes the instance to be restarted, at some point we'd like to move away from that model. If we were to evolve that capability how would you see that fitting in? Is this always going to get the list of nics via metadata or is it gathering that once at the beginning of time and caching it?

from rfd.