vg320 mgcp ports losing registration

[chuck47172]

CUCM System version: 12.5.1.13900-152

VG Show Version...#sh ver
Cisco IOS Software, VG3X0 Software (VG3X0-UNIVERSALK9-M), Version 15.4(3)M3, RELEASE SOFTWARE (fc2)

------------------------------------

Any idea what would cause MGCP FXS ports to go Unregistered and then back to Registered about every 30-40 seconds?
I have two VG320 and one VG310 exhibiting this behavior. After a cold start the problem stopped happening on two of them but the third VG320 continues to exhibit this odd behavior.

Here's something else I noticed where "NotifyRequest" counters keep incrementing and "IP address based Call Agents statistics" keep failing...

VGM01492#show mgcp statistics
UDP pkts rx 275750, tx 277323
Unrecognized rx pkts 0, MGCP message parsing errors 0
Duplicate MGCP ack tx 0, Invalid versions count 0
CreateConn rx 26, successful 26, failed 0
DeleteConn rx 6, successful 6, failed 0
ModifyConn rx 20, successful 20, failed 0
DeleteConn tx 13, successful 13, failed 0
NotifyRequest rx 136275, successful 131433, failed 4842
AuditConnection rx 342, successful 342, failed 0
AuditEndpoint rx 130942, successful 130942, failed 0
RestartInProgress tx 6234, successful 6234, failed 0
Notify tx 1893, successful 1893, failed 0
ACK tx 262769, NACK tx 4842
ACK rx 8139, NACK rx 0
Collisions: Passive 0, Active 0

IP address based Call Agents statistics:
IP address 10.34.150.169, Total msg rx 275661,
successful 270819, failed 4842
IP address 10.1.110.36, Total msg rx 89,
successful 89, failed 0
System resource check is DISABLED. No available statistic

DS0 Resource Statistics
-----------------------
Utilization: 0.00 percent
Total channels: 43
Addressable channels: 43
Inuse channels: 0
Disabled channels: 0
Free channels: 43

[mradell]

First thing I would do is get that IOS updated on those gateways. A while back we had quite a few MGCP issues with our VGs and upgrading them all to 15.8(3)M6 solved all our issues and has been very stable. That said, the newest release is 15.9(3)M6, so you may want to try that as well. I'd also work to confirm there were no connectivity issues from a network perspective. An MGCP gateway sends a keep alives every 15 seconds to it's primary subscriber in the form of an MGCP NOTIFY. If it doesn't get a response from the sub it will send a few more and then contact the secondary sub if it doesn't get a reply. This will cause MGCP RTMTs and make it look ports are unregistering/re-registering. To see if that's been happening look at the output of "sh ccm-manager". You can also see this behavior, along with all MGCP signaling, by running "debug mgcp packets" and capturing the output.

[chuck47172]

Thanks for the reply.



I found what I believe are two bad cheapo garbage analog phones preventing MGCP from restarting.



I disabled both phones (shut the port) and then did another MGCP restart and now all of the other ports are staying registered.



Bad phones are likely the root cause of all of this.



If I see a port that looks like it is on a call (show voice call summary) I can run the "show voice call status" command and if I see no active calls I will shut those ports down and then physically check the devices.


#sh voice call status
CallID CID ccVdb Port Slot/DSP:Ch Called # Codec MLPP Dial-peers
No active calls found

[mradell]

It's possible, but just as an FYI, those stuck ports are one of the common issues we were having before updating the IOS on our VGs. After updating them all it stopped happening. We have about 200 gateways/4k FXS ports with a mix of 224s, 310s, 350s, and 450s, so i had a pretty good sample size to compare with.

[Elliot Dierksen]

I have seen issues with various flavors of analog MGCP VG's when the MGCP package configuration in the VG didn't match what CUCM thought it should be. That made the config download fail. I agree with @mradell that an older IOS might not play nice with newer versions of CUCM.