Choosing dial peers

WRKNonCCNP

I have 2 voip dial peers (using sip)on the same router set up to go toward two remote sites (across WAN). The session targets are set correctly. I am load balancing between the dial-peers using 'dial-peer hunt 1' and both dial-peers have a preference of 1 and the same destination-pattern parameters.

My problem is that when a remote site fails, the dial-peer going toward the failed site doesn't disable and therefore calls don't failover to the active dial-peer. How is failover and recovery accomplished between dial-peers? If one site fails, i want all calls to go to the other site.

Ideas anyone?

Network Diagram:

ACD===[R1]

---

{WAN CLOUD}

---

[R2]====ACD
____________________`

---

[R3]====ACD

R1 dial peers:

dial-peer voice 100 voip
destination-pattern 2...
session target ipv4:<ip of R2>
preference 1

dial-peer voice 200 voip
destination-pattern 2...
session target ipv4:<ip of R3>
preference 1

How do i failover if the session target of one of the dial peers is down?

tokhss

good question.. i would like to know too.

I am not experienced enough yet in this feild but reading through some articles, you may not be able to achieve a failover if both dial-peers have the same preference value.

I was also reading through some other material that stated a Gatekeeper was needed for that type of dial-peer failover.

Bah.. hope someone has the answer.. i gotta start reading the CVOICE already.. hopefully this stuff is in their.

tmlerdal

Preference was my first thought. Could it have something to do with the route not aging out fast enough?

WRKNonCCNP

Thinking about it further, it seems that there isn't a way to accomplish both load balancing between the dial peers AND failover. Having two dial peers is like having two static routes. You can load balance across them if their AD is the same, but you never know when the next hop used by one of the static routes is down. I guess that is why dynamic routing protocols were invented and probably why i should be using some sort of sip server/client architecture instead of static dial peers.

I can think of doing it with IPSLA probes and EEM scripts to remove the dial peer config when that dial peer's session target is down, but that is slow and still drops calls. Or using tunneling with keepalives to determine when a peer is down, but otherwise I see Call Manager Express in my future!

Luckycharms

Well there are a lot of different ways to accomplish the intended outcome.
First lets address the configuration you have posted, it appears that dial-peer 100 and dial-peer 200 are h.323 dial-peers. If this is not the case and the configurations has been edited please post the entire configuration for proper evaluation. (Anonymised obviously) Also would you mind posting a debug of the ccsip messages for true evaluation of the failure of the dial-peer voice. This will tell us if it is truly trying to roll over the call and the timers are failing it or if you are correct in the fact that the hunt stop is some how begin invoked and stopping the call.

debug ccsip messages

( Be careful if you have alot of call's going this will produce alot of output!!!)

WRKNonCCNP

The dial peers have "session protocol sipv2" configured. i found a workable solution that uses routing in the WAN cloud to provide redundant paths to the same ip addresses (R2 and R3 both announce the same pair of loopback IP addresses, but one is preferred from each router). Half of the sip calls get dropped until convergence occurs, but once it does, all calls go towards the available router. Cool.

Luckycharms, thanks for that command. I had missed that one somehow and could have used it for the past few days.