If the link status could be propagated somehow, that would be nice. Perhaps if 
the VSW driver could be polled for link status from the physical link and the 
vnet driver could probe for that status. Obviously, we wouldn't want the vsw to 
go down or for guest domains on the same physical server to not be able to 
communicate with eachother. So this would probably require some extra work and 
possible changes to IPMP when vnet devices are utilized. The logic would be 
like this:

If IPMP is enabled on a vnet interface, it probes the vsw for physical link 
status. If the status goes down, it initiates a failover of the VIPs to the 
other vnet interface in the group. 

If all vsws in an IPMP group report link_down, then IPMP keeps the VIPs up on 
the primary vnet interface until the status changes.

I have no idea how complicated something like that would be, but it would 
atleast give customers an option if LACP is not available.

Octave
--
This message was posted from opensolaris.org

Reply via email to