I've recently bought a CRS328-4C-20S-4S+RM with intent to use as a simple access layer switch - dot1q trunk in via one of the SFP+ ports, then client connections out via the SFP ports. Each client has a dedicated VLAN untagged/retagged at their respective port. The switch shipped with RouterOS 6.42.5, I upgraded to 6.43.8 during my testing, without seeing improvement to the issue I suffer from. Of course, I performed the firmware upgrade, too, including multiple reboots.
In my initial testing, I've been testing each function with the device running RouterOS, including port INGRESS and EGRESS rate, tagging/untagging VLANs, etc. As my test bench doesn't have any fibre capabilities at the moment, all testing so far has been using the copper RJ45 ports from the 4 combo port group.
During testing, I discovered that when changing the egress rate for a port (Winbox/Switch/Port) from "" (default, nothing) to anything, the switch behaves as expected (change from default nothing to 200M, traffic drops to 200M as expected) but when I click the little UP arrow to revert to no entry in the field, the switch locks up. This requires a reboot to resolve.
The same behaviour DOES NOT occur when doing this on port INGRESS. I can go from default to 200M (or any other value, it works fine) then back to default and the switch continues to function.
If I set egress rate from default to 200M (again, example rate) then back to 1G, the switch behaves normally - it's the act of clicking the up arrow to revert the field to default that causes the lockup.
Does anyone else have one of these to test this function with? I haven't tested this in SwitchOS, as I'd prefer run RouterOS for this application - I feel the SNMP stack in SwitchOS is missing too many tree entries to be deployed in my particular application (from experience with a CSS326).
I have supout files, a full log of my testing, exported configs incase anyone wants to duplicate my setup directly.