> > - Is the hypervisors supposed to have a special configuration, ie vlans > for specific tasks, or could it simply be a untagged LACP connection with a > single ip address ? > > > untagged is preferred. We also support tagged interfaces… we support > bonding; where we want to move to is to be able to support multiple L3 > interfaces without LACP. > > ​Found the above thread in Aug 2014 ...
1. Compute Node --> TOR Is vRouter able today to do L3 load balance across N physical L3 interfaces to TORs ? Outbound routing to remote NHs would be default to TORs from each such fabric interface. Questions: - What would be limit to such ECMP paths ? - How to detect TOR liveness (BFD, light down) ? - What hashing algorithms would be available ? I see number of deployments still uses native linux bonding + LACP, but is this really needed if we would have L3 ECMP out of the compute node ? 2. TOR --> Compute Node While we are at the TOR <-> Compute interface what is the most common way to inject compute NH at TORs into underlay routing ? Simplest would be static routes to loopback over all connected interfaces. Other then that any plans to make it lightweight and dynamic? Thx, r.
_______________________________________________ Users mailing list [email protected] http://lists.opencontrail.org/mailman/listinfo/users_lists.opencontrail.org
