Routing single local host internet traffic through remote IPSec tunnel gateway

possible a policy-(default) route would work.

i would try:  policy routes / from:special hosts / to:any(or better needed destinations) / services: any (or known needed) -> Gateway: IPS-Router

Hi dirkkotte, thank for your reply. I read in some other posts that you can't forward packets in VPN tunnels via Policy routing...

anyway I tried your suggestion because it was a worth and easy try but unfortunately it's seems that it doesn't works...

Ik would agree with dirkkotte that (policy) routing is probably the issue.

With (IPSec) VPN tunnels you can connect all local networks attached to both UTM's (provided that you enter them in the local networks list of the connection). However, access to the internet will still be done through the default gateway of each UTM, so your HQ machines will be routed through the WAN interface of the HQ, unless you redirect the traffic of the specific server(s) trough the tunnel to the WAN interface of the other UTM.

Instead of policy routes / from:special hosts / to:any(or better needed destinations) / services: any (or known needed) -> Gateway: IPS-Router  on the HQ UTM as dirkkotte suggested, I would try  policy routes / from:special hosts / to:internet / services: any (or known needed) -> Gateway: IPS-Router on the HQ UTM as this would limit the route to traffic going to the WAN interface instead of any.

You mentioned that it seems that this doesn't work. Can you elaborate on how you came to this conclusion, eg traceroute ? 

Also important would be that either the SNAT rule on the branch UTM would have the 'automatic firewall rule' checked or, preferably, you would create a firewall rule on the branch UTM allowing the desired traffic to go out.

If Policy routing wont go through the VPN tunnel, I believe static routing (gateway route from special hosts to IPS-Router) could also work. In this scenario you would have to create the necessary firewall rules.

Regards,

Karl-Heinz

Hi Karl,

thank you for partecipating and for your kind reply.

I tried several policy routes as you suggested, both with Group of specific hosts as well as a single test host. When I activate the policy routes the internet services goes down, destinations become unreachable. Tracert become unreachable as well at first hop. It seems that for any reason packets can't be routed into existing VPN tunnels.

On the other side, as I said to Bob in the above post I can use remote gateway using proxy service and in this way at least internet browsing route through the needed route.

Don't know how to implement in the correct way so that whole internet traffic route through vpn, and this only for a specific subnet or Group of hosts...

Thank you !

MR

Hi Marcello,

Hmm, I'm curious as to your exact configuration. As I am afraid that this question may come up in my configuration in the future, I wanted to know for sure why things did not work or how to get them work.

By the power of virtualization, I've created a test setup. This setup is very simple, but I think it represents your question.

I've installed two UTM's with two interfaces, one WAN and one internal.

Both UTM's have the same default gateway in this setup (I've only got one internet connection here at home), but I don't think this influences the result.

My simple setup for UTM-1:

I created a firewall rule any <->any

I masqueraded the internal network to WAN:

Now I've got a Windows server on the internal network, ip address 192.168.5.10, gateway 192.168.5.1, dns 8.8.8.8. I've got another server, ip address 192.168.5.20 gateway 192.168.5.1, dns 8.8.8.8.

this works, eg tracert to this forum:

So this is your HQ server so to speak.

Now I've got a second UTM, UTM-2 with this simple setup:

I created a firewall rule any <->any

I masqueraded the internal network to WAN:

In this network, Ive also got a server ip address 192.168.6.10, gateway 192.168.6.1, dns 8.8.8.8

this also works, eg tracert to this forum:

I've created a simple IPSEC connection between the two UTM's:
First gateway definition on UTM-1:

and gateway definition on UTM-2:

In this example both gateways are initiate connection, but this is not mandatory.

Then I create the connection on UTM-1:

and on UTM-2:

Enable the connection and I've got a VPN between the two UTM systems:

I can test this by pinging the Windows host on the other end.

Without VPN from Host 1 (HQ):

with VPN from host 1 (HQ):

So now back to your question, how to route traffic from Host 1 (HQ) through UTM-2 to the outside world.

As shown before, traceroute from this host goes through UTM-1.

I've created a gateway policy route bound to the Intern interface of UTM-1, with the HQ host as Source Network for any service going to IPv4 internet to be sent to the Intern interface of UTM-2 :

If I enable this rule and I do a traceroute from Host 1 (HQ), I now get a different path:

As you can see the traffic is now  routed through UTM-2. 

 The same traceroute from the other host in the HQ network that is not part of the gateway policy route shows the 'normal' path:

As my own 'home' UTM is the gateway to the internet, I checked the results in my firewall log by going to a simple website with one ip address to see if the source address did change. Strangely enough it did not, although a TCP dump from the console on UTM-2 did show that the traffic was going out of UTM-2 and not UTM-1.

In the end I did something counter-intuitive and created the following SNAT rule on UTM-2 (your branch office so to speak):

This seems to work:

Remember, in both cases, the HQ machine (192.168.5.10) was the one creating the web request.

So the external address from UTM-1 was SNATted to the external address of UTM-2. This one I cannot really explain, but probably this is due to the fact that both UTM's have an external address in the same subnet. To be sure I would have to test with two separate external addresses. I haven't got access to such a config yet, but will try to get it. As soon as I do, I will post the final result.

So according to this simulation it should be possible to route one specific host (or network group of hosts) through a VPN tunnel to an external address.

I don't know if this answer will still be helpful for you, as Bob has provided a solution that has proven to work in your situation. But it has been fun figuring this out and hopefully someone can use this info :)

Best Regards,

Karl-Heinz

Hi Karl-Heinz,

thank you so much for your real commitment to the post :-)

More or less our environment is likely your virtual test environment, except we have hardware appliances and two different internet connections.

here is our tunnel 

and I can ping both gateway from each site as well.

I have a test host WS10 that can browse internet:

So I create Policy route:

When active internet become unreachable:

Even with SNAT rule on remote site:

Tried with automatic firewall rules and by check "Rule applies to IPSec Packets" box

same result ... Internet Unreachable

Seeing this I have to agree with Bob that for some reason policy routes do not apply to IPsec tunnel...

Hi Karl-Heinz,

thank you so much for your real commitment to the post :-)

More or less our environment is likely your virtual test environment, except we have hardware appliances and two different internet connections.

here is our tunnel 

and I can ping both gateway from each site as well.

I have a test host WS10 that can browse internet:

So I create Policy route:

When active internet become unreachable:

Even with SNAT rule on remote site:

Tried with automatic firewall rules and by check "Rule applies to IPSec Packets" box

same result ... Internet Unreachable

Seeing this I have to agree with Bob that for some reason policy routes do not apply to IPsec tunnel...

Hi Marcello,

Some questions just to be sure (since we now have both a working policy route through IPSEC and a non-working policy route through IPSEC):

Is the RemoteIPSEC gateway the LAN/Internal adapter on your branch UTM? It's the only difference I can think of for now between your environment and the VM test environment

Can the WS10 host ping IP adresses in de branch LAN if the policy route is not active? In my test environment I accepted the Remote LAN for each connection and offered the local LAN.

Regards,

Karl-Heinz

Hi Karl-Heinz,

yes to both questions, UTM are gateways of the sites and I can ping host in the other site from any of the hosts, when the route policy is active the host become unresponsive..

Regards

MR

Keep in mind the security associations here, you are not doing host -> ANY in the ipsec tunnel but specific networks / hosts to eachother. 

IPSEC would drop all traffic that doesn't match the security association. 

You can get around this for inbound traffic by doing a Full NAT on the remote gateway that changes the source to go through the tunnel and hit the site on the other side. 

Something like:

For traffic from ANY host

Using service: whatever service

Going to: External of UTMgateway2

Change destination to:  Webserver IP across the tunnel

Change source to: Interface IP of this side of the tunnel

Check rule applies to IPSEC packets. 

You guys might both be interested in considering Sophos UTM multiple S2S IPsec VPN mit Failover – Tutorial (DE).  All of the screens are in English, Marcello, so even if you don't read German, you should find the article accessible.

Cheers - Bob

Just a quick update on how I work arounded finally the question. I created a RED tunnel and bridged a new interface in HQ UTM to it. Then I connected a little router WIFI to the bridged interface. When people from branch office come to HQ with their laptops, they connect to the bridged WiFi so they can continue to access internet presenting Branch Office WAN IP,  mantaining their accesses on cloud resources.

A big thank to all, every ideas and suggestions came from this community 

Just a follow up on this as I am setting up a similar setup right now - Do i need to actually have two separate tunnels/gateways/connections created? I'm not understanding how this segregates the traffic whether they are listed as separate networks within the tunnel or separate VPN connections/tunnels? 

Not sure what you mean by "a similar setup," Aaron.

Cheers - Bob

Similar setup in the sense that I have a branch office and a HQ and trying to route all the internet bound traffic for a subnet to another site for an exit point.

I tried with the policy based route which breaks the connection and doesn’t work / nothing pings or routes.

If I try adding 0.0.0.0 to the IPSec tunnel - the tunnel doesn’t establish.

My question I poorly worded last night was - do I need two entirely independent tunnel connections or just the subnets to be listed separately in the local/remote networks on either side respectively ?

Show us pictures of the Edits of the IPsec Connection and Remote Gateway for both sides and tell us if/where Web Filtering is being done - the exit point site or not or both.

Cheers - Bob

Well, now I can get the tunnel to establish (I believe it was a double-NAT issue on the "Branch" side). My goal here is get only one specific subnet to route its internet traffic via a second ipsec tunnel.

Unfortunately, when I get the tunnel to establish, all the branch side sophos traffic (as in, traffic generated on the Sophos itself) seems to try and go through the tunnel or just quits working altogether. Traceroute no longer works, cannot ping externally, etc. A review of the routes under Support>Advanced shows that the Sophos is trying to default all traffic via the new tunnel.

This isn't the intended routing so first and foremost I need to find a solution to this. 

The branch Sophos has 1 NAT rule that routes internet bound traffic from a source of "any" out on a WAN public IP of the branch Sophos.