04-04-2022 12:07 AM
So we seem to have a bit of mystery over here.
We have several rendernodes attached to a Cisco SG350X-24, which in turn is attached to an SX350X-12 through the 10Gb ports. Attached to that SX350 are also 2 data serverss (Windows Server 2016 OS) through a 10Gb port on an Intel X722-DA (not sure if DA-2 or DA-4).
We also have several workstations with the same issues that most likely go through another SG350X-24 at one point.
Writing data (from rendernode or workstations) TO one of the data-servers is just normal expected speed, so the issue is ONLY on receving data from data-servers.
At first we thought auto-negotiation was the suspect, but after setting the all 10Gb ports on the path to 10Gb Full-Duplex, nothing improved.
So, we're looking for a Sherlock Holmes here who can help us out, because at this point we are at a loss as to how we can solve this. None of the settings we changed on either of the switches or the network card seemed to do anything.
04-04-2022 02:14 AM
- On the SG , check port counters for the windows server(s) , watch for errors if any , on windows use eventvwr , look for networking errors , or check interface statistics such as counters on the windows server too (if possible).
M.
04-06-2022 12:43 AM
Hi thanks for the reply,
for now, we seem to have narrowed the issue down to the Windows Server 2016 10Gb signal being converted/translated to a 1Gb signal on a Cisco switch (be it the SX350 or the SG350)
So next course of action is to update the switches to the latest firmware, and after that probably upgrade the servers to windows server 2019 (as none of them seem to display this issue)
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide