Endpoint Dropouts from TMS

[philiprogers]

Hi folks,

Wierd query.

I am working with a user using TMS 13.1.2 and a mixture of endpoints(MXP - C Class - Ex 90) and VCS's MSEs

They have started to experience a lot of systems losing HTTP connection to TMS.  They drop out for a few minutes only and then re-connect.

Location makes no difference, nor is there any pattern on time of day, nor of system type.  It's totally random.

Cisco have checked the TMS config and looked at Wireshark traces - can't find anything

Network team cant find anything, neither can the users server team.

Anyone out there got any ideas or come across this before.

Cheers

Philip

[philiprogers]

Meant to say 438 occurances in the last 24 hours

[epicolo]

Hi Philip, what do you mean when you say "loosing HTTP connection to TMS"?

Where do you look to collect this statistics?

Are you using TMS appliance ?

Every 15 minute TMS starts the "TMSDatabaseScannerService", the next scan start 15min after the last one.

If a system is unavailable, the service will get that status until the next scan, or until the endpoint sends a trap to Cisco TMS

PS: 100 systems can take up to 30min to retrieve info.

The HTTP feedbacks are sent from devices whe some event occours (and is configured).

The TMS can also send SNMP to quickly discovery if the system is reachable or not at the specifed interval.

Regards

[philiprogers]

Thanks Elter - a couple of answers below. 

Regards

Philip

Where do you look to collect this statistics?  TMS is set to send an email if an endpoint does not respond.

Are you using TMS appliance ? No, Virtual Server

[epicolo]

So, you got an e-mail every time the TMS try to contact the enpoint and it does not respond.

Is the HTTP enabled on endponts? Are you using SNMP to quickly discovery accessibility?

You can try to reach the endpoint using the browser directly on the TMS server, to check if the conenction is ok.

Ex: http://A.B.C.D/status.xml

If ok, you can capture (sniff) the TMS port to check the TCP hadshake initiated from TMS and the HTTP get sent to endpoints to check if you got an answer or not.

I hope it help.

[philiprogers]

Thanks again, the reason it is weird is that it is totally random, sometime it works, sometimes it doesn't.  Always manages to re-connect and different sets of endpoints.  No correlation to time of day either.