Trying to establish a highly resilient edge for a client who desires to spend little coin.
The Internet edge consists of 2 NSA 2600's in an HA deployment.
The Inside edge is a pair of CS-3850-24t's
Each 3850 connects to each SW
3850 G1/0/19 to SW1 x0
3850 G1/0/20 to SW2 x0
3850 G2/0/19 to SW1 x2
3850 G2/0/20 to SW2x2
All 3850 ports are set as access ports in the same vlan (800)
The 3850's have 2 additional downstream 2960 switchstacks connecting to it
All LAN SVI's are on the 3850
The active switch is switch 1
If I power off switch 1, internal LAN routing and intra connectivity fails over to switch 2 ie if I have a continuous ping going from devices on different vlans and switches, there is a 3 tick no response then it replies again. That's acceptable.
However, a continuous ping from inside to an internet host dies during and after failover. My perspective is that the primary SonicWall has the former active 3850's mac address (from G1/0/19) in it's ARP table and unless I recycle that primary sonicwall, there is no egress or ingress connectivity. If I do recycle the primary sonicwall, I eventually reestablish complete connectivity.
Am I on the right path that it's some kind of ARP issue at the sonicwall? If so, what are my options (not a sonicwall dude)?
If not, what other options do any of you suggest to create an unattended failover environment?
I've considered creating routed links rather than l2 connections between the switch stack and sonicwalls with OSPF????
Thoughts?
Thanks for any help
