Ethernet connection gets lost after some minutes

Hello,
we have problems with touchpanels, based on T20 Colibri. We have many devices at different customers that are connected to the customers LAN over managed switches with VLAN configuration. In our WEC7 there’s no VLAN enabled on AX88772B. The devices worked without any problem in the network for more than 2 years. Since November 2020 the following network issue appears:

After switching on the devices, you can ping them on the network for about 4 - 6 minutes. After that time they are not reachable any more. You can’t find them on the switch at the port and they aren’t in ARP table of the pc that pinged them any more. After resetting the devices or unplug and plug in the network cable they work again for about 4 - 6 minutes.
Customer says that nothing has changed in network and software of the switches. The ports of the switches have been changed with no result and we exchanged our devices but the problem is still there.

Does anybody has an idea what this could be?

Thanks, Dirk.

Hi @VNA

We never saw this kind of behavior, but we also did not test much with VLANs. Is the VLAN switching only done on the external Switch, or do you also configure something on the units?

I would try disabling the VLAN to see if it’s related to that.

Or test with another device instead of the T20 unit in the same network

The 4-6 minute timeout also does not seem to have any explanation on the device HW or SW side.

The strange thing is also you said it was working until now and without changing anything suddenly it starts behaving strangely? Maybe there is some other network component which was changed or auto-updated in the meantime?

Sorry for the late answer,
we configured a VLAN and I can’t get the problem here. It’s just at the customers network. We have 6 T20 units there, all with the same image version and the same software version of our software. All of them are connected to the same switch in the same VLAN. The switching is all done on the external switch, there are no tagged packets coming to the devices. 5 of 6 T20 Devices have this strange behaviour, one is running without any problem. The customer says there was nothing changed or updated.
At the moment it’s difficult to visit the customer.
I’m just examing a wireshark protocol of all network traffic to one device.

Hi @VNA ,

Is the problem staying on those 5 specific modules even when swapping the port on the network switch with the one module that works? that would be interesting to know. Maybe the problem follows the port on the switch.

Other than that i really have no idea as i never saw network problems that start after minutes of boot.

Ah… one last thing… are the modules still fully functional apart from not reacting to pings or did they maybe crash completely? Or maybe there is no way to tell as Ethernet is the only external interface?

Hi @ germano.tx,
after doing many checks in the network environment of out customer now we have found the problem. After disabling IPv6 on the Colibri devices the devices stay stable in the network. It’s a problem with icmp6 neighbor solicitation in the image based on BSP 1.4 (beta3 release candidate). We have 8 devices in the network, 2 of them running an image based on BSP 1.3, they are running without a problem. The 6 devices with the image based on BSP 1.4 are always loosing connection after 5 minutes. We logged the network traffic with wireshark and found out that this always happens after a neighbor solicitation packet. It seems that on the customers site there came up a new device with IPv6 enabled which is sending the neighbor solicitation and since this time we had this problem.