I just purchased a new CSS326-24G-2S+RM to evaluate. For the first step, I updated the firmware to v2.10. While I'm not here to critique the Web GUI, I will assert that it looks more like "proof of concept" code than a commercial production release.....
I was able to setup a test config with multiple VLANs, including VLAN trunks and untagged ports and tested the SFP+ interfaces with 10GigE DAC cables. Note that currently I'm only using 802.1q VLANs. RSTP, LACP, IGMP, and EVERYTHING else is set disabled.
This all is stable and works fine, as long as the network is lightly loaded. Once I heavily load the network (particularly with heavy duplex traffic), I start having odd traffic-flow pauses, with hosts becoming completely unreachable. I've tried every combination I can think of in the config setup. Nothing resolves this. This problem isn't related to using the SFP+ ports. It happens just the same using only the 1GigE ports.
As a final test, I setup only two 1Gige ports, isolated from all other ports (none of which were connected). The first port accepts only tagged VLAN traffic, the second port untagged on the same VLAN. The tagged port is plugged into a backbone switch (tagged uplink). The untagged port is plugged into a Raspberry Pi4 (MTU 1500), which was handy. I NFS mount a network share on the RPi4 and left it all morning with a script copying a couple small files back and forth. No problems at all....THEN, I start a linux kernel compile (make -j 4). Within 1 minute, the NFS share is completely unresponsive (host won't even respond to pings), as seen from the RPi4. While this is all happening, I'm connected to the RPi4 via an ssh session (which remains responsive, no problems) from another system connected via the same backbone switch. SSH continues to work fine.....
As the next step, I simply unplug the 2 network cables from the CSS326 and plug them into a preconfigured DGS-1100-08 switch (it was available for testing). NFS starts working perfectly, including "torture" testing. Swap the 2 ports back to the CSS326, it again works fine until I heavily load the network and the problem is back. Swap back to to the 1100-08, working perfectly again.....Note that I'm not rebooting ANY hardware---just moving network cables.
Again, note that I've tried flow control off and on (CSS326, backbone router, RPi4). This same behavior is seen between PC hosts (with intel NICs). It's not the RPi4 causing this!
Anyone have any ideas??? I'm starting to pull my hair!