Dual WAN failover and partial failure

[pschwendiman]

We had an interesting situation after an electrical storm moved through where a few of our dual WAN clients didn't have any Internet connectivity.  Generally speaking, they are on a wireless broadband and something like DSL/cable/T1.  One customer in particular has a cisco 1941 router setup for dual WAN which tested fine during install (pull either connection, external IP changes dynamically and no difference is noticed by user).  Well, this storm knocked the wireless broadband out enough that it wasn't usable, but would respond to pings randomly.  Because the wireless is the fastest of the two connections it has a higher priority but because it was down but still responded to some pings, the traffic didn't go through the DSL that didn't go down.  Any suggestions on how to make the dual WAN more reliable in these partial-down situations?  I thought about playing with the network service detection (we have a customer on an RV042 who experienced the same problem) but am not real sure what I could change that would make the connection more reliable, especially on the 1941 router.

I don't expect this to happen a lot, but my customers weren't very happy that they are paying for dual ISP so they don't loose connectivity and then do loose connectivity.

Thanks for any suggestions!

[Calin C.]

Hello,

I would say:

- interface tracking with route changing - only works if the interface really goes down

- IP SLA - monitor the remote end (e.g. with ping or some tcp packets; the frequency has to be at about 1-5 seconds, if you encounter big problems with wireless)

- EEM - change from active to back route after a certaing message appear in the logs (like icmp unreachable, interface down...)

The idea is that there is no best solution in my opinion. The best solution would be the one that suite your environment best, but for this you have to do some tests and also check what features your Cisco device support.

The above mentioned are just an example.

Regarding your wireless issue. I also had this behavior some time ago. I shutdown device, completely disconnected from power source (including external power adapter if it has one), wait for about 5 minutes to completely discharge the electrical charge and then plug it back. It worked for me. I could shutdown completely the sites affected, because this were branch ones and not so critical to network environment.

Cheers,

Calin