There have been many times we have to setup devices and the device software uses a discovery tool to auto detect for initial configuration. For example our security cameras, we have a dedicated vlan for which spans many switches/buildings. To use the camera discovery tool we have to use the tool from a machine that is on the same switch as the camera we are setting up. Not a big deal as we have gotten by this way no problem. Once the camera is given and IP we can place it anywhere in our network on that vlan and it works fine.
Well we now have a situation where the software has to use auto-discovery every time it starts in order to detect the devices and display them within the software. Works great while all cameras are on the same switch but as soon as we place a camera in another building the discovery doesn't work. I don't know much about the software but we are trying to get information from the vendor on this. We are able to IP these cameras off the single switch. When I move them to another switch I can ping and browse to it (HTTP) but the software will not pull it in. There is no way to direct the software to the IP either, it has to use it's built in discovery to find the cameras.
Can anyone clue me into what might be going on here? All of my searches are coming up with nothing. I'm thinking I just need to know the right thing to search for but am clueless :(
Info of network:
Catalyst 2960x's distribution switches
Catalyst 3650 doing the routing (even though routing is not coming into play here)
Hello @mikeschriver ,
you need to understand how this SW discovery happens.
For this purpose you can use a SPAN monitoring session with source port = a video camera port and destination port = a port where you connect a PC with wireshark running to perform packet capture.
The only suggestion is to verify if you use VTP in this case if VTP pruning is enabled for the video CAM Vlans the unknown unicast MAC addresses are not flooded over trunk ports.
In other words if the server remembers the camera MAC address and attempts to send a frame with a unicast destination = Camera MAC before the camera has sent a frame that frame would be treated as unknown unicast and should be treated as a broadcast.
If the downstream switch has an empty MAC table for the CAM Vlan it sends VTP prune messages upstream to avoid to receive frames.
This does not happen when all devices are connected to a single switch in this case VTP pruning has no effect.
If all your switches are in VTP transparent mode try to perform the packet capture and to get information from the vendor the discovery may be using some form of multicast and you may need an IGMP querier to make IGMP snooping to work well (but this would be valid also for a single switch scenario so it is just an example)
In the past I had a case where a vendor of medical appliances was using a non IP proprietary protocol but sending the frames over dedicated to IPv4 multicast MAC addresses this triggered IGMP snooping. To make it work after packet capture we had to disable IGMP snooping for that specific Vlan as being not IP no device was answering to IGMP queries causing the switches to block this traffic treating it unwanted IPv4 multicast traffic.
Hope to help
First thanks for your reply. I believe you are onto something for sure. I did a packet capture and can see the main tricaster server sending multicast packets.
So to answer some of your points above all our switches are on a VTP domain with manual pruning. I believe this solves any auto pruning issues since I could not ping the cameras across the network until I manually allowed that Vlan over each trunk port.
I did find this information about the vendors autodiscovery... https://support.newtek.com/hc/en-us/articles/218109477-NDI-Discovery-and-Registration
So the discovery is using multicast as you suspected (mDNS). Now I just have to figure out how to configure things so these multicast packets reach the devices. I have some learning to do here so if you know anything that may speed up my learning curve I am all ears :)
I'm thinking I need to make some settings on our layer 3 switch so IGMP works differently on that Vlan?
Again thanks for you pointing me in the right direction! I'll post more if/when I resolve this.
Hello @mikeschriver ,
from the document you have linked:
The mDNS Ethernet frame is a multicast UDP packet that broadcasts to:
This is good news as IGMP snooping does not block link local multicast addresses 224.0.0.X.
So this should work even with IGMP snooping enabled and without a querier.
To have a working setup of IGMP snooping you need either:
a router or multilayer switch with multicast routing enabled with PIM enabled on the Vlan and IGMP to act as an IGMP querier or a switch that is configured to act as the IGMP snooping querier that is to be a subset of a multicast router and to send IGMP general queries on the VLan so that all switches can use IGMP snooping to create correct L2 distribution list of ports for each multicast group.
if each camera sends a multicast stream out of 224.0.0.X you need the querier/ IGMP snooping querier
be aware that VTP auto pruning is actually a different thing then the so called manual pruning.
Configuring manually the list of allowed Vlans on a trunk limits the number of STP instances running over it ( if using PVST or Rapid PVST).
VTP auto prune refers to suppression of BUM traffic for Vlans permitted over the trunk where the downstream switch has no clients over it. VTP auto prune does not limit the number of STP instances running on the trunk and on the switch.
Hope to help