PBR/NAT mechanics

[ClintL]

Good morning everyone,

I have a weird issue that I think is related to how PBR or NAT works and/or something we need to program differently on our Palo Alto. I’ll try to explain our setup and the issue that occurred.

We have our main circuit that is utilized for VPN tunnels, NAT forwarding rules to our servers and pretty much anything that is not standard Internet traffic. We’ll call that Circuit A.

We have another circuit used for mostly standard outbound Internet traffic for certain VLANs. This is also the circuit that our DNS servers use for requests. It is high bandwidth but it is not a business class circuit with SLAs so it is not used for critical applications. We’ll call that Circuit B.

Like it was stated above, all NAT forwarding rules use Circuit A with the exception of one server (let’s call it Server 1) where we have a NAT forwarding rule that uses Circuit B. Note that this server is in a VLAN that uses Circuit A for outbound traffic.  This server also has several internal IP addresses associated with it and they all use Circuit A for outbound.  The NAT forwarding rule only points to one of these internal IP addresses.

Last night the ISP did planned maintenance on Circuit A which took it completely down. When this happened the NAT forwarding rule to Server 1 (which again goes over Circuit B) stopped working. My current guess is that somehow even though the NAT forwarding uses Circuit B to Server 1 it somehow uses Circuit A for outbound on this rule maybe on one of its other assigned internal IP addresses. I didn’t think NAT would work like that but nothing else comes to mind as to why the NAT forwarding rule to Server 1 over Circuit B would stop working. Any thoughts? Hopefully my explanation on our setup wasn’t confusing but if clarification is needed please let me know.  Thanks!

[reaper · Accepted solution]

Hi

Are your NAT rules set up including a destination interface?

you may need to set the interfaces so NAT can't 'travel' between interfaces

I'm guessing the NAT rule itself may not have stopped working, but routing may have been an issue, it's possible the returning traffic from the server is not hitting the PBF to circuit B properly and is actually egressing circuit A

return packets will also be inspected by the pbf table and then hit the routing table if no matches are found

if the returning packet doesn't match any bpf configuration it will go out the default static route

you can try setting up a specific PBF policy for the server so it's returning packets are sent over the pbf, possibly you may need to enable symmetric return

regards

Tom

[ClintL]

That was my theory but you laid it all out better than I could explain it.  I'll make those changes and test it out.  Thanks Tom!