Devices Unregister on CM

[Neily]

We are having lots of problems with devices unregistering on our call managers.

Even the MOH server (which is on the call manager) unregisters!

We get various reasons including: TCP-Timeout; Failback; Phone Keypad; CM-aborted-TCP

Has anyone got any ideas as too why?

THANKS

 

[jneiberger]

Hard setting only one side is horribly bad. Never--ever--hard code one side and leave the other to autonegotiate. Always leave both sides at AUTO unless you have a very good reason for hard setting them. If you must hard set them, do so on BOTH sides. Even then, you're asking for trouble. It is an unreliable setting.

I've re-read the thread and I don't see where Neily mentions using non-Cisco switches, but that's irrevelant anyway. If you're using switch made in the last few years then AUTO is the best way to go. Only hard set the settings if you are experiencing errors or duplex mismatches due to faulty NIC drivers.

 

[pndscm]

I just did a search on Cisco.com and we are both right and both wrong. Depends on how you look at it.

http://www.cisco.com/warp/public/473/46.html

Neily mentioned non-Cisco switches at 10:03 post today. At least when he typed 'We are keep being told its a 'Network Issue' but as we don't use Cisco switched they won't help us!' I assumed he mispelled 'switches' to 'switched'.

 

[jneiberger]

On that page, notice the following note:

"Some third-party NIC cards may fall back to half-duplex operation mode, even though both the switchport and NIC configuration have been manually configured for 100 Mbps, full-duplex."

I opened the TAC case that got that added to the documentation. Cisco originally did not have that language in the documentation but they finally agreed that I was right and this caveat needed to be added.

 

[pndscm]

Hard coding both devices is a valid way of checking for connectivity problems per this same document. Let's agree to disagree so we can focus on the main issue of Neily's problem. Let's see what he comes up with when he does his testing tonight.

 

[jneiberger]

Agreed. It will be interesting to see what he turns up. I have seen weird connectivity issues with bad code on switches before, but I won't be able to help much with that if he isn't using Cisco switches.

 

[whoisg8ts]

This is a known issue and it is resoved by upgrading the driver for the "Adaptec Embedded SATA HostRaid" controller to version v6.0.82.0

We have installed CM on a cisco 7825 Server (Which is really an HP platform) and we had this exact problem. It was discovered that a housekeeping task in the device driver (specified above) resulted in a buffer clear on a regular basis. (every 5 hours) This interupts data flow to the hard drives and results in the CM failing to respond to the MGCP NTFY keepakuve messages.

Hope this helps.

 

[Neily]

I was just reading through this and saw that I never updated when I said I would!!

Anyway, all cabling checked out ok.

I never found a solution and haven't checked recently to see it is till happening!