We’re running 11.2.1HF3 with a bunch of virtual servers configured with their pools for nPath. It works well except for one virtual where autolasthop was mistakenly disabled; the traffic flowed as expected except during failovers - then it fell down (firewall in front of node dropped return traffic as out-of-state). Turning autolasthop back on made traffic flow again during failover.
When the node returns traffic to the client via its router, and later comes back from the client to the VIP as part of a normal TCP conversation, the F5 is able to forward the traffic back to the same node even though the connection table lists ServerSide IP as the vip (instead of the node as with SNAT or AutoMap).
I’m suspecting the F5 → node flows continue until FIN because the F5 makes use of the autolasthop database to find its way back to the node chosen during load balancing. Does that sound right? Is there a way to inspect the autolasthop database?
Thanks in advance!