MGCP - ISDN - CUCM registration problem

[1977bjorn]

Hi all!

We have an open case on this, not sure on the progress so I thought i just ask here to see if anyone have seen this before.

In short, we have a cluster of 4 cucm - recently upgraded to 8.5 (we saw the problem in 7,5 as well but not in 6.1). I do not think this is dependent on cucm-version but thats just a guess. Sometimes out gateways (cisco 28/29xx with isdn pri E1-cards, ios 15.x.x.x) that are mgcp enabled seems to loose their registration. The problem result in the isdn message "channel not available" seen in the "debug isdn q931". If you try to reset the gw from the webinterface from the publisher the gw does not react... if you go in to the webinterface for the cucm that the gw is registered to and do a reset from there, it works. Often the gw works ok after this, sometimes you have to reload the gw in order for it to work again. If we prior to this do a "sh ccm-manager" it is often registered to the backup ccm - not the primary! (no network issues - we have checked)

It is a complex problem and we do not see any obvious hints in isdn/mgcp-debugs. If we do a redundance test and for instance bring down "side a" by blocking the traffic - the re-registration works fine.

We have seen the problem in production, often caused by a wan outage (often very very small drop on the wan links as we have dual providers and routing with very fast convergence). So it seems like minor drops cause more problems than just blocking  site completely. I know there are a lot of variables in here, so this is mainly to see if there is anyone else that hase seen the same problem. My guess is a some bug in cucm or ios...

Anyone?

regards

Björn

[paolo bevilacqua]

Why MGCP that is source of limitations and problem

H.323 or SIP would give you a much more robust and felxible solution.

[Jefferson Islan Silva]

Could you please reply us with the router config?

[1977bjorn]

Hrmm, sorry but no, I´m not allowed due to policy...

As I said, there is no problem with the mgcp-config - everything is working perfectly from time to time... There is no single point of failure either, so a configuration file would not say that much. This worked in the 6.1 version of cucm and ios 12.2.24(T4). We are now on 8.5 and ios 15.x.x.xsomething. I suspect either ios-bug of some kind (but not likely) or some cucm-cluster issue that relates to mgcp operation. I mean, the main issue is that the isdn operation as well as registration fails during a failover - and that is not consistent either.

We have tac working on it, so I´m sure we will see a solution, I was just curious to see if anyone have seen the behaviour before...

[1977bjorn]

No doubt that H.323 or sip would provide a simpler solution seen from a technical perspective. We do have this as a long term plan, but there is also an enterprise contact center behind this solution and a huge dialplan also involving precense servers that handles sip transport. So, that is not happening right now anyway :-)

[acampbell]

Bjorn,

If as you say you cannot rest the MGCP gateway from the pub server I think you should be looking at replication issues in your cluster.

Look at the database summary on RTMT to start with.

Regards

Alex