Forum Discussion
andy_12_5042
Nimbostratus
NimbostratusSNAT AUTOMAP
I understand how autonat works but I cant seem to find this answer. If you are not using floating ips and and have multiple public ips assigned to egress vlans, how does the map work.
What I mean by that is it possible that there will be some randomness in the address used to automap. I am thinking that this is a wrong design choice in cases where your application source ip needs to be a constant. In other words I cant have random source ips coming back to clients as I am in a hosted services environment.
Can anyone give some logic here as to how this would work with the scenario I gave above and if it makes sense to use automap where I need the src to be same for clients every time. Also if I add self-ips, my understanding is that it will also use the new ones as that is the concept of automap and why you have endless ports for nat to use....
- Chris_Miller
Altostratus
If you're talking about inbound traffic, you're going to map to a self-ip on whatever VLAN faces your pool members. Do you have multiple self-ips facing your pool members? 
hoolioCirrostratus
Hi Andy,
If you have multiple IP's in a SNAT pool or multiple self IPs on an egress VLAN with SNAT automap, LTM will round robin through them. I don't think there is any way within default configuration to change this behavior.
You can use an iRule to select the SNAT address based on the client IP. Here is an example in the post below. You could also modify this to check the selected destination server IP in LB_SELECTED using [LB::server addr].
SNAT pool and persistence
http://devcentral.f5.com/Forums/tabid/1082223/asg/52/showtab/groupforums/aff/31/aft/813179/afv/topic/Default.aspx813195
Aaron
andy_12_5042Ok thanks that it what I was looking for.
