Intermittent ARP/client-to-client communication issue on Smart Hub 2 (Wi-Fi to Wi-Fi only?)

[nickymoo]

I'm trying to work out whether this is a known Smart Hub 2 issue or whether I should be looking elsewhere.

My network is:

BT Smart Hub 2
Pi-hole providing DHCP and DNS
Raspberry Pi 5 (Ethernet)
Raspberry Pi 5 1GB (Wi-Fi) - 192.168.1.12
Raspberry Pi 5 1GB (Wi-Fi) - 192.168.1.13
MacBook Pro (Wi-Fi)
iPhone (Wi-Fi)
iPad (Wi-Fi)

Every few days one of the Wi-Fi Raspberry Pis becomes unreachable from all Apple Wi-Fi devices.

For example:

MacBook to  Pi51 - nothing
iPhone to Pi51 - nothing
iPad to Pi51 - nothing

However, at exactly the same time:

Wired Pi to Pi51 - works fine
Pi51 to MacBook - works fine
Pi51 to Internet - works fine
Tailscale to Pi51- works fine

So the Pi is definitely still alive and connected to Wi-Fi as it connects flawlessly over tailscale from any device (when Smart Hub 2 is out of the chain)

Diagnostics

When the fault occurs:

On the Mac:

sudo arp -d 192.168.1.12
ping 192.168.1.12

The ARP entry remains incomplete and the ping fails.

Running tcpdump on the Pi at the same time:

sudo tcpdump -eni wlan0 'arp or icmp'

shows no ARP request arriving from the Mac at all. 

The Pi does still see broadcasts from other devices and can ARP wired devices normally.

The interesting part is that if I manually add the Pi's MAC address on the Mac:

sudo arp -s 192.168.1.12 88:a2:9e:e8:8c:09

everything immediately starts working. Ping works. SSH works. HTTP works.

So ordinary unicast traffic is fine; it's only the ARP discovery that's failing.

Things I've ruled out

Reconnecting Wi-Fi on the Pi doesn't fix it.
Reconnecting Wi-Fi on the Mac doesn't fix it.
Pi-hole doesn't appear to be involved because this happens before DNS is ever used.
The Pi itself is alive throughout.
The problem affects multiple Apple devices, not just one.
It has happened with more than one Raspberry Pi on Wi-Fi (although I normally access the second one over Tailscale, so I notice it less).

It feels like the Smart Hub is intermittently failing to forward ARP broadcasts from Wi-Fi clients to one wireless client, while still forwarding normal unicast traffic.

I know 100% that Pi51 is sitting on the internet and working perfectly as it's connected to a ADS-B receiver and feeds into FlightAware and FlightRadar24 with a 100% uptime reported by both services. And I can always get to it via Tailscale it's just I can't access it locally via the same network.

I turned off DHCP on the Smart Hub 2 a few weeks ago thinking that was it, but direct pings show ARP misses and it didn't help the situation. Rebooting the SH2 every day doesn't help either.

Has anyone else seen behaviour like this on a Smart Hub 2?

I was thinking of buying a BT Business Hub off Amazon or Ebay and using that, but I have BT Digital Voice and I understand that the land line isn't enabled by this route. I was also considering taking the SH2 out of the picture entirely and replacing the router, but I have the BT Hybrid Connect device which I imagine won't play nicely with anything but the Smart Hub 2.

Also MAC address randomisation is turned off. When the issue occurs, nothing can speak to it via the network (iPhone, iPad, MacBook Pro). The SH2 is running the latest software version. FTTP. Excellent wifi coverage throughout the house.

[licquorice]

The implication of your post is that you normally leave DHCP enabled on the BT hub. That most definitely needs to be disabled if you are using PiHole for DHCP and DNS. You can't have 2 DHCP servers on a network. 

Try a factory reset of the hub and then disable DHCP.

Neither Digital Voice or hybrid connect can be used with anything other than a Smart Hub2 or residential Smart Hub 3.

Your hybrid connect device is pretty superfluous with fttp which is unlikely to suffer the same degree of line faults as a copper connection.

[nickymoo]

---

Thanks. I turned off DHCP on the BT Hub about three seconds before I enabled it on my wired Pi 5 4gb. 

Will try a hard reset later if no other solutions turn up.

hybrid connect does get used in the wee hours several times a month when my exchange seems to take stuff offline. It is useful and I’ve got a vulnerable adult here who really needs the lights to turn on when she speaks to them so I would like to keep it.

if I replace the router/hub entirely with anything else then I guess I completely lose my land line?

[licquorice]

---

@nickymoowrote:

---

if I replace the router/hub entirely with anything else then I guess I completely lose my land line?

---

Yes

[nickymoo]

Factory reset didn't solve it.

[licquorice]

It certainly is a strange fault. 

Have you tried using Wireshark to see what is actually happening?

Probably no consolation but I also use PiHole for DHCP and DNS and transmit FR24 data and haven't experienced any problems.

[nickymoo]

Haven’t tried wireshark yet.

It’s odd as the ARP call never reaches Pi51, but Pi51 is deffo online transmitting flight data and is 100% accessible via Tailscale which runs on all my Pi devices. 

I only moved DHCP to the PiHole to rule that out from the hub then realised the calls fail just on IP address.

But all works perfectly when connecting from Tailscale.

It’s been going on for a while and I had considered the Pi 5 1gb model might have some wireless defect but then Tailscale rules that out too. It’s just been bewildering.

I was considering moving from BT to EE to get a more modern hub but similar packages to mine are much more expensive.