Unraid loosing network connection

[BlindOracle]

Hello everybody,

lately my Unraid 6.9.2 server has serious issues with 10G network connections.

Looks like I can reproduce the issue, when I'm runnig a Cloudberry Backup, after about 1 - 2 hours running the network connection drops and the whole server including a running VM becomes unavailable.

 

VMs are using a secondary 10G uplink, the first NIC is for management and docker.

I already created a diagnostics report today, as soon as the link dropped again.

I already replaced network cables, verified that the switch ports are fine, plus I updated to the latest BIOS version.

 

My backup plans only include backups for network shares, so no wiered system files or such.

 

Thanks very much in advance!

 

Blind

[ChatNoir]

3 hours ago, BlindOracle said:

I already created a diagnostics report today, as soon as the link dropped again.

You should probably attach it so someone can have a look at it.

 

If the connection drops after heavy use, did you check that the NIC has sufficient cooling ?

These cards can run pretty hot and are generally passively cooled. That's fine for a server with ton of airflow. On a consumer grade chassis, it can be an issue at time.

[BlindOracle]

Oh very true... Haha

unraid-diagnostics-20220406-1208.zip

[JorgeB]

NIC is losing the link, not much you can do on the Unraid side:

 

Link detected: no

 

[BlindOracle]

Hi everybody,

it took me several days to figure out what my issue was. I’ll explain to provide a resolution for my issue. 
my Unraid server is running on a SuperMicro Mainboard X10SDV-2C-TLN2F. It has 2 10Gbe links and currently there seems to be a known issue with those adapters. In my case when both NICs are connected via 10g, that the system will drop network connections unexpectedly at various times. I found this article on the net, describing my issue, and I followed it, contacting SM support. 
https://tinkertry.com/how-to-work-around-intermittent-intel-x557-network-outages-on-12-core-xeon-d

 

SuperMicro provided me a firmware patch version sdv23a for those NICs, but applying it didn’t change anything for me. I received and applied a version sdv23b a day later and that worked well for me, with a little con, as I had to connect NIC0 to a 1Gbe switch. Then it looks very stable again. AlsoI tried connecting NIC0 to the 10g switch and limit the port‘s bandwidth to 1Gbe, but that didn’t eliminate the problem.