Frequent Ethernet failure on sbRIO-9607

[KHartley]

We are using a sbRIO-9607 and were conducting reboot testing with our system. We identified that the static network connection between the sbrio and the host computer was failing to establish about 4 in 100 power cycles. Our test ran for 800+ iterations with 40+ connection failures. Nothing is in the error logs that would state our application is failing to boot and the sbRIO has a character display off the FPGA to show the ip address which it retrieves during startup that is correct.

We had swapped out a USB to ethernet adapter we were using for the host computer and replaced it with a laptop with a built in network card and the same issue is appearing. Checking for frequency as we speak.

We have established that by using a network switch we can eliminate this ethernet failure but this adds $100 to our system for what we do not need. The issue can be resolved by unplugging the network connection but this is an industrial machine that needs high reliability.

Based on my recollection there was a known issue with HP switches that led to this issue. We are using a HP laptop or a Startech USB31000S adapter.

The network configuration of the sbrio is 

Static

IP 192.168.2.3

subnet 255.255.255.0

gateway 0.0.0.0

dns 0.0.0.0

Auto Negotiate

Interrupt

The network configuration of the laptop is 

Static

IP 192.168.2.4

subnet 255.255.255.0

gateway 0.0.0.0

dns 0.0.0.0

We are using LabVIEW RT 16.0 with CompactRIO 17.6 I've verified the safemode image 4.0f0. I remember this issue around the 16.0 release.

[a_clucker]

Hey Kyle,

Are you using runmode or safemode during this testing?

[KHartley]

Run Mode.

We just finished a long term test with the laptop that has an integrated network card and it failed 12 cycles in 479 iterations.

Our test details.

We cycle power using a daq card and relay to the sbrio. The laptop stays powered. 

We apply power after 15 seconds of remaining off and let the sbrio boot and run our application for 2 minutes. 

We attempt to connect to sbrio once at the end of two minutes before we power cycle again.

We have tried adding a repeat attempt to connect but typically the network is not functional until the network link is severed or the sbRIO is restarted.

I think Nathan Sullivan was testing this way back in 2015.

[KHartley]

We did additional testing overnight trying to set the negotiation mode to base1000 explicitly and polling on ethernet. Eth1 is disabled and not populated.

After 600 tests it failed to connect 20 times.

[a_clucker]

Hey Kyle,

Yeah, I chatted with Nathan after seeing your first post. I found the CARs from the 15.0/15.5 timeframe, as well as a occurrence of one in 17.0 with a powered ethernet switch. Part of the fix for that made it into 17.6 and it looks like the other part made it into 18.0. Can you run a test with 18.0 as a troubleshooting step?

The other possibility here is that this is a brand new issue we haven't seen before.

[KHartley]

I'll probably have to wait a few weeks before updating the driver software. We're not inclined to update the driver software quite yet as we're making bug fixes for shipping systems and do not want to add that on top.

Glad to hear there is a potential fix.

[KHartley]

We continued testing over the weekend. We dropped the baud rate to 100base to see if that would help. We saw reports of 4 connection errors over 1700 iterations. So its less prevalent but seems like its a low-level issue. We do not have many calls that could load down the ethernet port except for 2 which could spew about 1/3MB very rapidly a few times a second.

I'll see if I can set up a copy of my computer to try the later compactRIO driver.