Hi @Manned6141 

Thanks for posting in our business forum.

Manned6141 wrote

Hi there,

 

I have an issue that seems to have been brought up a few times by others, but there was never an actual resolution so I'm posting in hopes of finding one. I've seemingly tried everything I can think of, and the problem persists.

 

The problem:

10G SFP+ ports on both my TP-Link switches auto-negotiate to 1G speed when PCs and servers are booted up. Problem persists across all SFP+ ports on both switches.

 

Current workaround(s) to get 10G connectivity:

1. Reboot switch each time one of the servers or PCs is turned on; or
2. In either web administration or Omada
   • manually change port speed from auto (or 10G) to 1G
   • enable flow control
   • wait 5-10 seconds
   • save and apply changes
   • wait 5-10 seconds
   • refresh page
   • change port speed from 1G to 10G
   • disable flow control
   • wait 5-10 seconds
   • save and apply changes
   • wait 5-10 seconds
   • refresh page
   • repeat from step one if it doesn't take (usually does)

 

Equipment:

TL-SG3428X - V1.30 - FW 1.30.0

It appears that this is the only FW version available for the hardware revision I have, so I have not tried any other FW versions.

 

TL-SX3008F - V1.0 - FW 1.20.0

Previously tried FW 1.0.1, 1.0.4 as well as 1.0.5. All versions had the same issue.

 

3 x Intel X520-DA2 10G SFP+ NICs (1 in each server/PC all running various versions of Linux, PCIe 4 slots, plenty of available lanes).

Previously tried a Mellanox and HP 10G NIC, with same problem. Thought they were the issue so grabbed the Intel ones to test, but no change.

Multiple Linux kernels tested, as well as Windows. Problem persists.

 

Intel E10GSFPSR transceivers (1 in each NIC and SFP+ ports on switches, connected with optical cables).

Previously tried "TP-Link compatible" 10GTek transceivers as well, but the problem persists.

 

Omada Controller is running in a docker container on a local hypervisor.

 

Troubleshooting steps taken, none of which helped:

• Unplugged all ports, and tried plugging in devices one at a time
• Unplugged fiber from PC, booted, then connected
• Updated all systems and drivers to latest, also tried older versions to see if there is any change
• Pleaded with the devices to "just work"
• Factory reset all devices, set up in standalone mode
• Factory reset all devices, set up with Omada
• Set port profiles in Omada with explicit settings for 10G and Auto, tried combinations of different profiles between different ports (for example port 1 or 25 as 10G, 3 and 27 as Auto, etc), with various MTU and flow control settings.
• Threatened the devices that they will be returned and replaced with Unifi if they don't start working as intended.
• Tried other transceivers, and combinations of equipment I have
• Ran speed tests, and once the manual workarounds are done, I get awesome speeds, averaging 9.72 Gbps throughput between any and all combinations of the devices using iperf3.

 

The problem has persisted for over a year, when I only had the SX3008F in my setup, and I was hoping the new SG3428X I just picked up would fix it, but no luck. It is clear that it should all work, and all of the equipment is fully capable since it works after manually setting the port speed. However having to do one of the two workarounds, sometimes 3-4 times a day is incredibly frustrating.

 

What can I do to resolve this asap? I'm still in the return window for the SG3428X, but I'd rather get it to work if I can.

 

Thanks.

So in summary, the problem is it does not link to 10Gbps after the reboot of the switch with the PC with 10Gb NIC. Is that correct?

I have a question, you have used a 10Gbps transceiver? It does not relate to the problem here. The problem is the compatibility with the Intel NIC.

You are certain that you have updated the NIC to the latest? What's the version of your NIC now?

If you force the link to 10Gbps, does it work after the reboot?