Network connection to a single host randomly drops, need help diagnosing

[pdc]

I am having a problem with connectivity from a specific host (Surface Pro 7/Win 10) to another specific host (Raspberry Pi 4). My network is an AiMesh of 2x AC86 running the latest Merlin firmware (386.2_4). The surface is connected via wifi (typically to the main router) and the pi is hard-wired to the mesh node.

The Surface wireless adapter is Intel Wi-Fi 6 AX201 160MHz and has the latest Intel driver (22.40.0.7 3/3/2021). I have not noticed a problem between any other clients (a mix of Windows, Android, and iOS).

I typically notice this problem when I am accessing the pi via RealVNC from the Surface. At random times, but not very often (between twice a day and every few days), VNC will report the remote host closed the connection. The first time this happened I thought the pi had crashed, but every other client was still connected.

When this happens I can still access the main router from the surface, and from the main router I can still ping the pi. Other clients can also still access the pi.

If I try to ping the pi directly from the surface, I get a variety of "Destination host unreachable" and "Request timed out". And then after 30-60 seconds it reconnects.

About the only clue I have is "arp -a" on the surface shows the pi no longer listed, but I have no idea why that might be.

Any tips on how to diagnose this further?

 

[Justinh]

I had what might be the same issue between two Windows hosts. To solve it, I ended up editing the host file on the source machine (the Surface in your case) to map the remote host name to its IP address. (I use the router to reserve IPs, so both hosts always have the same IP address.)

I never did figure out why these two hosts had this one-way problem. I could ping from the second machine to the first, then the first would be able to ping the second by hostname. After some time, the first could no longer ping the second using hostname and I'd have to repeat. Thus, the host file solved the DNS issue.

 

[eibgrad]

@Justinh suggestion for a local host files makes sense *if* the problem is actually w/ name resolution. But the OP wasn't clear if he could eventually ping based on the hostname, ip, or both. If it's an arp problem, a local hosts file isn't going to help. You might need to add a static arp entry for the RPi to the Surface machine. Doesn't explain *why* there's a problem w/ arp, but if that fixes it, then clearly something is amiss at the arp level.

 

[pdc]

For my scenario the hostname resolution says no such host, so the pinging was done using IP address. I tried adding a static arp entry, but it says permission denied, even with a shell running as administrator. Maybe because there's already a dynamic entry? I'll poke around some more with google...

 

[Justinh]

For my scenario the hostname resolution says no such host, so the pinging was done using IP address. I tried adding a static arp entry, but it says permission denied, even with a shell running as administrator. Maybe because there's already a dynamic entry? I'll poke around some more with google...

DNS is not resolving the hostname. Modify the hosts file. Why would you mess with ARPs when it is a DNS issue?

 

[pdc]

DNS is not resolving the hostname. Modify the hosts file. Why would you mess with ARPs when it is a DNS issue?

Because pinging by IP address doesn't work either, I get a variety of "Destination host unreachable" and "Request timed out", and then after 30-60 seconds it reconnects. It is odd that DNS isn't resolving either, now that you mention it...

I will try the host change. Of course now the problem hasn't happened in a while so maybe some update has fixed it (hah, wishful thinking )