I have 2 RH based servers in different datacenters and use GRE tunnels to send ranges from one to the other.

The issue I'm running into is that when I restart the network to bind the tunneled ranges it takes an average of 4 seconds per IP. (Around 16 minutes for a /24).

I'm not sure if the issue is that by default the server is trying to ARP the range on the local network and since it's not setup in the router it has to wait the default amount of time for a response before timing out or if it's because the default on the system is to restart the network adapters first and then any tunnel devices?

I could use some help figuring out the best way to solve the issue, I've considered:
-setting the ARP time out lower (if it would help and I can find out how to do this)
-using a shell script to bind the IPs without restarting the network (which I've tried but the supposed universal method for doing this on 2.4+ kernel versions does not seem to work).