I have an Ubuntu 18.04.5 LTS, system on an AMD 64bit system with 16 GB. Recent Zoneminder install has cranked up traffic. But I did not use this system much prior to the ZM install.
After a while (minutes to hours), the system stops communicating. When this happens, the ARP cache is mostly showing incomplete in the mac entries, and the only way to reset is to reboot or unplug the Ethernet. Eventually it happens again.
I have tried 3 different drivers up to and including the latest from Realtek. Currently, using the r8168, 8.045.08, 4.15.0-118-generic, x86_64 driver per dkms. No real diff between them. The systems I am trying to ping are all in the same local subnet. Interestingly, the default router, and other IPs that are not sending or receiving traffic, retain their MACs in the ARP cache UNTIL I try to ping them or otherwise communicate with them. So once this starts, every ARP entry will move to “incomplete” until no comms with anything is possible. To be clear all IP and ping traffic immediately fails, but the ARP cache doesn't start to show Incompletes until I try to send or receive to each one. It makes sense to me why I can't communicate. The system does not know the MAC!
The question is why is ARP failing. Multiple other systems on the same local IP subnet have no issues communicating with these exact same external systems all on the same local switch. I have tried different switch ports, but after briefly correcting itself, it just happens again. I have no duplicate IP's. It is like ARP validation just goes to sleep along with periodic ARP flush. I am not sure what to look at next.
Any ideas why it keeps dropping? I see similar questions like mine here, but none say it was ever resolved and trying some of the suggestions did not help me.
UPDATE******
After many failures. I decided on a complete reinstall of Lubuntu 20.04 LTS. Sadly, I have the EXACT same problem. It looks like it uses the same driver also. tcpdump shows my system is constantly trying to ARP to refill the incompletes in the table. I can also see those broadcasts hitting the network on another machine. Once the problem starts, the problem box never sees any replies (in tcpdump), and all of the ARP entries age out . And just as before, simply unplugging the ethernet cable for a second, or a reboot fixes the issue. Using the gui to disable/enable the interface does NOT fix the issue. And by "fix", I mean temporarily. Eventually the problem always comes back. So essentially the "receive" on this systems goes to sleep until I physically reset the interface somehow. Transmit is working fine.
Ethtool shows the same driver of 8.048.00-NAPI for the r8168. My bios says the chip is 8111G. Based on what I found, the driver is the same for 8168 and 8111.
So I have ruled out at least one rev of the OS. That leaves the driver or Ethernet hardware? Or perhaps a config issue the crosses these to versions of the OS??
Sure could use some suggestions here. Thx.
Update*****
After seeing another post where they had a similar issue, which was solved by a bios update, I checked mine again. I am on the latest bios for the MSI AM1I MB, which is V10.2. It is from 2014 but no updates have been made available.
Update 10-30-20******
After reading another post, I tried forcing the interface to 100Mb FD. That just created another issue, a periodic full system freeze requiring a reset button or power cycle. Not even the sysrq worked.
Fortunately, My USB Ethernet arrived today. So I tried the new hardware and it's automatically downloaded driver. Sadly the "packet receive stop" happened again!! Exact same symptoms. After running fine for 5min to an hour, suddenly and randomly, no packets are received per tcpdump. Outbound (transmit), is working as before. ARP packets are being broadcast, other machines are responding, but this system "sees" NO packets coming in. In all these tests and versions, the ethernet maintains link and the activity LED flashes as if nothing is wrong.
Now keep in mind that my "load" on this machine is 5 IP cameras feeding the Zoneminder app. It keeps a fairly high receive load on the ethernet interface.
That leaves just a few things in common, the fact that I have installed Zoneminder on bother versions of Linux, and some of the hardware, not including the Ethernet components.
I am starting to wonder if I am the only one using Ubuntu anymore, or if this is a low use forum. I have not had a single reply since I fixed my original typo. Do the Devs even look here? I must admit I expected a few responses after so many days.
