Replaced an old unmanaged switch under a ControlLogix, 1756-L81E with a 1756-EN2T, for an IT-supplied Cisco IE3300 last week. Hanging off it are two 1794-AENT, three 1734-AENTR and four PowerFlex 525.
Every time that switch reboots, and IT rebooted it twice doing firmware, the I/O is fine for about five minutes and then every rack drops 16#0204 inside the same minute, drives included. Kepware on the same VLAN never drops, which is the bit I can't explain.
On a hunch I turned IGMP snooping off and the drops stopped, but IT put it straight back on because it's in their standard config. Swapped the EN2T for a spare and that changed nothing either. So it's a switch setting, I'm just not sure which one. Any ideas what to ask IT for?
Kepware is TCP, class 3, it doesn't care. Your I/O is multicast. Two things: is there an IGMP querier on that VLAN, and what's the snooping group timeout. Ask IT for the output of show ip igmp snooping, they'll have it.
> is there an IGMP querier on that VLAN
IT says "the router does that". The VLAN is layer 2 only, no SVI, nothing routes it. The show output says Querier: none. Group membership timeout 260 s, though I'm not sure that's the default.
Querier none is your answer. Snooping listens for the IGMP joins and only forwards multicast to the ports that joined. Those joins get refreshed when something sends a query. No querier on the VLAN means no queries, so after the 260 s timer the switch prunes every group and the adapters stop hearing their producers. Five minutes after each reboot, every rack, 16#0204. Snooping off works because it floods the lot, same as the old switch did.
The fix is a querier on that switch, one per VLAN, ip igmp snooping querier with a spare address in it. Their standard assumes a router on the VLAN and there isn't one. There's a writeup with the multicast bit here: https://plctr.com/advanced-plc-networking-and-communication-protocols/
IT enabled the querier on VLAN 30 with 10.30.0.2 this morning and rebooted the switch twice to prove it. 40 minutes, nothing dropped. In TIA world this wouldn't come up at all, PROFINET RT is unicast. I'm not calling it fixed until it's run a shift.
Watch the 525s. They're first to drop when a group gets pruned.
Two shifts, nothing dropped, so I'm closing it. IGMP snooping with no querier on the VLAN, the switch pruned the multicast groups after the 260 s timer, about five minutes after each reboot, and ip igmp snooping querier on VLAN 30 sorted it.
What I still don't know is why it never bit before. The old unmanaged switch had no snooping, so it just flooded everything and nobody noticed, I guess. Thanks petek, northbay.
Get the querier line into their standard config while they're still listening. The next switch will have the same gap.