Support Forum
The Forums are a place to find answers on a range of Fortinet products from peers and product experts.
btp
Contributor

IPSEC tunnel does not re-route

I have a setup as image indicates. I have a main and backup tunnel in my VPN VDOM, that work as expected. When main link is unavailable, the 4G backup takes over. When wan1 is OK again, the traffic moves back.

 

Then I have another tunnel from root VDOM that connects to my management network, to ease management independent of which IP address the spoke might get. This tunnel also works fine, but it tends to get stuck over the 4G link - even though this route is less preferable than the main link. The routing table looks something like this:

 

(VPN) # get router info routing-table database

(...)

S       0.0.0.0/0 [10/0] is directly connected, IPSEC_BACKUP inactive, [5/0]

                  [10/0] via 31.149.50.53, wan2, [10/0]

S    *> 0.0.0.0/0 [5/0] is directly connected, IPSEC

     *>           [5/0] via 37.248.6.246, wan1, [5/0]

C    *> 10.4.13.0/24 is directly connected, internal

C    *> 172.16.100.0/30 is directly connected, mgmt-link0

C    *> 172.16.100.1/32 is directly connected, mgmt-link0

If I flush the tunnel, it moves to the right interface.

 

Is there any setting I might have forgot, or is this as expected?

-- Bjørn Tore

-- Bjørn Tore
3 REPLIES 3
Toshi_Esumi
SuperUser
SuperUser

Routing is not the issue as you showed that the backup is not selected as the best based on the distance 10. The issue must be related to the active sessions running on the backup IPsec when the primary came back. Since the backup is still up there is no trigger to remove the active sessions and re-establish them.

You can try like dial-on-demand over 4G or set up a routing protocol over both IPSecs, or shorten session timers. You need to match both sides not to have asymmetric routes.

btp

Yeah - well; a reboot fixed it. But I have seen this behaviour on an handful of muy devices - which is quite annoying.

-- Bjørn Tore

-- Bjørn Tore
btp
Contributor

So - for the interested: the problem was not the CPE at all. But the receiving end of the IPSEC tunnel. Since I use a loopback interface there, the unit did not process this correctly. Turning off NPU offloading for phase1 on that end solved it.

-- Bjørn Tore

-- Bjørn Tore
Labels
Top Kudoed Authors