I have a network based on 21 x Gbit Layer 2 Managed Switches, effectively in a star layout with one switch providing the links for the other 20 switches. All the switches are interconnected using VLAN Trunks. Each port on the network is individual set to provide access to one of 8 VLAN’s. The primary feed switch is hooked up to a pair of MikroTik routers using VRRP for fail-over. Each VLAN exclusively uses its own IP address range i.e VLAN 2 = 10.1.2.X, VLAN 3 = 10.1.3.X and so on. The netmasks are all set to 255.255.255.0 as each VLAN requires its own exclusive broadcast domain.
The question I have is quiet simple. When two machines on the same VLAN are trying to communicate this should occur across the switches and traffic should not go via the router.
However I am unsure what happens when two machines are on different VLAN’s does the traffic go via the bridge in the router?
I seem to be suffering from intermittent UDP packet loss around my building and want to eliminate the routers. I haven’t done an iPerf test yet (as most of the hosts I am having issues with are embedded devices - things like h.264 video encoders), however the common factor in the errors I am experiencing are that they are using UDP to communicate with each other.
UDP will show errors where TCP won’t because TCP will do retransmit until the data is complete. You would see that in a Wireshark trace if it was happening. You can do a packet sniffer trace on the interface where the “bad” network exists on the Mikrotik router if you have enough disk space. Open that in Wireshark.
Yes, anytime packets go between networks a router must be involved. You didn’t tell us how the routers are attached (individual cables per vlan, or a trunked port), so that link or links becomes suspect. You said something about a bridge but you should show your config or diagram it so we can understand it.
If it’s only a problem between 2 particular networks, or sourced from the same switch(es), concentrate on that. It could be cables, or cable routing. Or WiFi issues if you are using WiFi.
Do the “low hanging fruit” stuff first. Check for port errors /interface ethernet print stats-detail. Move from one port to another. Change cables. Etc.
MY Routers are connected via Trunks. Its all a bit odd because this all started last Sunday afternoon, when nobody was in the building - so no equipment was added or removed. Its been working for at least 4 years. Done all the basic stuff (cables, rebooted everything, etc). The odd thing is it ONLY seems to be happening between hosts on VLAN2 (10.1.2.225 > 10.2.2.230) so thanks to your reply I can basically exclude the routers and go an annoy people on the TP-LINK forums. LOL
Its also been getting increasingly worse, there are about 400 messages going from the controller to the device per day. When it started 1 or 2 messages a day went missing, now its 350+ per day. The controller is an Old 2008 Mac and the device is a Blackmagic Enterprise VideoHub (288 x 288 SDI Video Switching Matrix). Unfortunately the controlling software wasn’t my choice, and because of stupid licencing conditions I can’t upgrade the Mac or use a PC.