Dual ISP with DNS servers connected to one router over BGP, how to implement redundancy if one link or dns fails

[husqvarna1990]

Hello i am having real trouble configuring a network with 2 ISP routers connected to one router, each ISP router has a DNS server + ntp connected. I need to test if the server fails or the link to the server fails, then connect to ISP2 instead thus implement some redundancy. I am running OSFP and OSPFv3 in the client network and BGP to connect the ISP's. but i have not added a way to favor the route to ISP1 yet.

Will just adding the weight command enable me to favor a ISP ? 

I have though about an SLA to test if the DNS server is alive, but to be honest that is beyond me, i can do a basic ACL but have never done a route-map - sla.

Have i got the right idea of using the weight command in BGP to favor a ISP and do i need an SLA to test if the DNS server is still alive ?

Please help, i need to get better and your info shall help me lots over my career :)

I have designed it in packet tracer as we use the same router and i can just copy the running config over via putty. So i have some pics and a file too of my current progress.

All i need to do now is get the Admin router to favor ISP 1 and to test if the DNS server fails and thus using ISP 2 and its dns server. Please help me.

Someone deleted my post yesterday, i just want to learn, please don't be harsh :(

[Richard Burts]

If you have 2 ISP connected to a single router in your network then favoring one ISP is easy and using the weight attribute is the simple solution and it should work fine. If you are running BGP with both ISP and using weight to favor one then failing over to the second ISP if the first one fails will happen automatically. That is one of the things that BGP does very well.

Adding DNS into the discussion makes it not simple any more if you want to change routing decisions based on whether the DNS at the preferred ISP is working or not. First you need a way to check on whether the DNS server is working. Then you will need some logic to detect the failure and then trigger a change in the routing logic. Probably something like an EEM script would do that. But that is fairly complicated logic.

HTH

Rick