My understanding of Arista's implementation of MC-LAG is relatively weak, so its likely I've misunderstood a deliberate limitation of how the
Taking a basic example,
switch1: vlan10-svi: 192.168.0.1/24 switch2: vlan10-svi: 192.168.0.2/24 switch3: vlan10-svi: 192.168.0.3/24
So in a given scenario of switch3 pinging each of the SVI's on switch1 and switch2, I'd expect to see,
# ping 192.168.0.1 80 bytes from 192.168.0.1: icmp_req=1 ttl=64 time=0.338 ms # ping 192.168.0.2 80 bytes from 192.168.0.2: icmp_req=1 ttl=64 time=0.338 ms
Whereas in reality,
# ping 192.168.0.1 80 bytes from 192.168.0.1: icmp_req=1 ttl=64 time=0.338 ms # ping 192.168.0.2 1 packets transmitted, 0 received, 100% packet loss, time 0ms
It is also the same when attempting to ping each respective SVI from switch1 <=> switch2.
If the either port on the portchannel on switch3 is shutdown - then reachability of each respective switch is possible - but it is never possible to ping both switch1 and switch2.
Now, I understand that VARP is likely the tool used to allow reachability of both devices (and in fact, does); but in the example where each switch needs to be independently accessed (eg. for management outside of
ma1) - it doesn't appear to be possible.
- Is this simply L3 limitation/design of Arista's MLAG?
- Should it be possible to ping between the mlag peers (not using the mlag peer link)
- Should it be possible for switch3 to be able to ping both switch1 and switch2