Can you connect to each device's non-floating IP? If so, that's a good bit of data, and you're probably dealing with ARP issues - especially if you've been doing fail over testing with frequent fail overs. The reason why is that BigIP issues a gratuitous ARP for all floating addresses (and virtual servers). In certain situations devices may not honor the GARP update and you've got bad arp caches laying around the network. Try this:
1) Confirm you can log on to each unit's self-ip address (.101 and .102 in your case).
2) If you can, great. Go to the active unit via SSH and run 'b load', which will force another gratuitous arp out onto the network and will hopefully coax a few more systems to update their caches. If this doesn't work, you may want to clear the arp tables on your switches.
If you can't get to the self-ips at all, then you're probably dealing with other problems.
I hope this helps.
-Matt