I have a weird situation with multicast that I hope someone can shed a light on.
I have a server directly connected to a 3850. It serves as command and control to several clients. For my test, one of the clients is directly connected to the 3850 too. The server is on VLAN A and the client is on VLAN B. When the server first comes up, it sends a multicast discovery to 239.0.0.5. The client replies back unicast, and the two exchange several messages to establish a session.
The server can either be configured to automatically send these discoveries, or it can be done manually. When we do a manual discovery, everything works. No issues. When we do an auto-discovery, it doesn't work.
A packet capture shows that in the case of the auto-discovery, the server is actually sending duplicate discovery packets for some reason. The first hits the switch with a TTL of 16. The second hits a few milliseconds later with a TTL of 15. Clearly there's an issue with the application. We have people working on that.
But, my question is, why does the switch drop the first discovery packet? Packet captures taken at both ends of the exchange show that it never makes it to the client. Therefore, the switch has to be dropping it. Theoretically, it doesn't know that a dup is on the way, so it should forward the first message. I've compared the manually produced discovery with the automatically produced one line by line, and there is no material difference in the two. Any input would be helpful. Thanks.
