UDP Connection Issue via Procurve 2650...

[Andy Piatt]

This has been an interesting issue...

Three proprietary devices with static IP addresses communicate via UDP in a server-client-client configuration. With all three connected to a consumer 8-port switch, they work without any problems. However, once we connected them to a working procurve 2650 switch, they could no longer communicate. I can still ping each of the three devices but they don't communicate with each other. Now, when the server device was connected back to the consumer switch and that consumer switch was then connected to the procurve, the clients and server could once again communicate (the client devices are still connected to the procurve). There are also no IP address conflicts on the network. Any ideas?

[Mohieddin Kharnoub]

Hi

Its an interesting issue really, the 2600 turns to firewall :)

What kind of communication is happening between the clients and the server,
Is it a Unicast with that UDP port number ?
or its a broadcast with UDP number.

Of course any router or Routing switch blocks any kind of UDP broadcast between different networks and thats why there is a way to enable some specific broadcasts based on UDP port numbers.

In your case, i assume that the 2600 switch has zero configuration and the Server and the Clients are in the same Vlan.

Did you try to run a sniffer to see whats going on on the 2600, and where the session is dead or timed out or .... ?

By the way, whats the UDP port number ?

Good Luck !!!

[Andy Piatt]

The communication happening between the clients and server is apparently broadcast. Though neither their documentation, nor the support staff can/will answer that conclusively.

Our 2650 has no special configuration so all clients are on the same VLAN.

This was tested locally and deployed remotely (I wasn't there in person). I have walked them through the drills that I would normally take in diagnosing a problem, but nothing as in-depth as a sniffer. For that, I will need to head there myself. My guess is that the server is continuing to broadcast, but the data is lost (or stifled) in-route. Moving the cable from 2650 to consumer 8-port brought the clients immediately back online. Moving it back to the 2650 showed the connection between client and server drop back out.

Again, all three still ping when connected to the 2650, but no displayed interconnectivity between them. I had a thought that the auto-negotiation between server device and 2650 is having problems, but the consumer 8-port is able to act as an intermediary between them? Sound plausible?

Support was only willing/able to tell me that the UDP port number used would be "above the normal range" which I interpret as >1024. They may use different ports depending on their configuration at the factory so, in their defense, they may not be able to tell.

Thanks for the help!