Hi,
Please be patient and understand this is not a conventional enterprise setup. My skill set is probably half way CCNA.
We have a large scale audio video system (massive multi-way divisible conference room) with Crestron NVX multicast video traffic (IGMP v2), close to 200 endpoints when all connected. It is completely isolated and self contained, with an optional portable "recording system" that is only plugged in as needed. That's where the trouble is.
The core is a stack of 7 C9300-48p. Switches were configured with a help of a CCNA (or likely higher) but I understand the configuration for the most part and have no problem CLI adjusting settings etc. In fact I have tweaked it a lot past initial setup and it is working flawless by itself. The portable recording gear has it's own C9200-12 "edge switch". That recording gear has the option to work on it's own, but when brought up to the conference room, some of the NVX devices remain connected to the edge switch, other get connected directly to the core.
Here is the problem, when the edge is connected to the core, the multicast traffic on the core stops for 20 seconds (during topology change I assume). This is not acceptable as it will interrupt on going sessions held in part of the space. So I discovered that if disable spanning tree on the edge switch - the mcast traffic no longer interrupts during connection of the edge switch to core switch (what we need). The switch will be out of sight and no easy access to, so I have no problem leaving STP off. However, in 50% of the live connections, the edge switch gets flooded with multicast traffic for about 1 minute. Past that point, everything works great, but I don't feel comfortable leaving it like that.
Setup details:
Single VLAN, but I have setup the uplink port as trunk, 10Gbps copper port
Matching configuration on the EDGE switch, via FS SFP+ 10Gbps copper
I did set the STP root priority as primary on the core and as secondary on the edge - no change
IGMPv2 Querier is set on both core and edge (cause edge needs to functions on it's own when not connected to main system), but the edge switch has a higher number IP address (lower priority), thus auto elect happens correctly and the core switch becomes querier to the edge when both connected. I did notice that with STP off on the edge, the querier change (auto elect) happens much slower than when STP is enabled, and I assume this is the reason for the flood. I did play with port settings (port fast, native vlan vs non native, access vs trunk, IGMP timings) to no avail so far.
I do realize someone will ask for configs, and I will provide as needed, but didn't want to clutter the initial post too much. Hoping someone has run into this and could point me to correct configuration, perhaps a different type of SPT or something of that sort.
We normally use Netgear AV line switches, and I did reach out to someone at Netgear, who confirmed what we need works with their switches as long as root priorities are set.
I appreciate you patience if you read till the end.
EDIT:
Someone will probably mention VLAN routing at some point, due to other details I don't want to confuse with it, it is not on the table at the moment.
