Hi,
Finally our IPv6 bug has been confirmed, and Avaya says it's fixed in both 5.1.2 and 7.0,
and if we want to test.
The only reason I'm considering release 7.0, instead of upgrading our 5.1.0 to 5.1.2,
is the fact that Multinetting is supported since this version.
For us as a hosting provider, it gives us many advantages because we had to give customers
that needed multiple subnets ... an extra vlan. And since we don't want to provide customers a trunk port (if I want to re-arrange my vlans at one point or
do something else with it ... they have to make a change exactly that time as well)
So what are the experiences with 7.0, if any ?
It's an ERS8610 with dual SF8292 with SuperMezz, and 2x 8648GTR copper blades.
Also ... the IPv6 bug is as follows :
We have 2 (Cisco in this case) L3 devices, each one connected to a single copper blade.
So very simplistic :
Cisco1#Portchannel1 => MLT#1 (LACP, vlan5)
Cisco2#Portchannel1 => MLT#2 (LACP, vlan6)
ERS86000 is configured with 2 IPv4 and 2 IPv6 static default routes towards the Cisco's.
We have ECMP enabled, so both IPv4 default routes are used.
(ECMP is not used for IPv6 in release 5.x, not sure how this is in 7.0)
And here is where it goes wrong.
You can ping every IPv6 host internally that's 'behind' the 8600 (eg where the 8600 is doing
the IPv6 routing for), but you can't reach (yeah, only 1 packet each 10 minutes or so) any host
where the IPv6 default route has to be used.
If we set a static route to some network that lives behind the Cisco's, and for all
the subnets that are in between ... it works.
Solution : turn of ECMP, or ... just give one _IPv4_ default route a higher metric/cost (also
effectively disabling ECMP)
It also doesn't matter whether you have one or multiple IPv6 default routes.
For some reason, it's trying to utilize the IPv4 default routes instead of it's IPv6 default route
when IPv4 ECMP is working.
