How to diagnose a mesh WiFi problem?
How to diagnose a mesh WiFi problem?
Author
Discussion

clockworks

Original Poster:

7,671 posts

174 months

Friday 21st August
quotequote all
I've been having problems with various smart home devices dropping offline - lights not getting turned on by switches or motion sensors, temperature sensors not reporting, etc.
I use Home Assistant, and I can see the various devices going "unavailable" for anything from a few seconds to a couple of minutes.

Problem got much worse yesterday, now doing it several times an hour.

I've narrowed it down to a wifi problem, as Matter over Thread devices are fine.

I've installed Ping app on an Android tablet, and I can clearly see very high pings (300 or more ms) every couple of minutes, plus packet losses that can last a few seconds. This is when I'm pinging the router.

The complication is that I've got multiple TP Link Deco mesh devices, mostly using an ethernet "backbone", to provide the 2.4 and 5ghz WiFi networks. I can't be sure what "route" the tablet is taking to reach the router.

Note that I get the same high pings and packet losses if I ping the two closest Deco devices.

Before I start taking the network apart, any ideas how I can pin down where the fault lies?


arandomstranger

76 posts

114 months

Friday 21st August
quotequote all
Is it possible to see what device is connected to which AP? Maybe they're not connected to the nearest one for whatever reason.
Check the 2.4ghz band is set to 20mhz rather than 40, as most IoT devices won't work properly at anything over 20.

ARH

2,050 posts

268 months

Friday 21st August
quotequote all
Check out how many devices your wifi mesh can cope with, Many ISP supplied routers can't cope with more than 20 or 30 devices reliably. I have no idea on mesh devices and their capabilities. This often shows up as devices just not working well although they appear connected.

This can be easily???? solved by running your own DHCP server rather than letting the router or mesh do it. When i had this problem I moved over to Pihole and setup up its DHCP server.

There are also other possible issues like, poor wifi signal, neighbours wifi on the same channel, and probably many other things that can interfere with the signal.

clockworks

Original Poster:

7,671 posts

174 months

Friday 21st August
quotequote all
In theory, the TP Link Deco app shows what clients are connected to each mesh device.
Problem is, it often doesn't show reality, as it seems to update the network map extremely slowly.
It tends to show clients as "connected" even though have been powered off completely for many minutes. I basically can't rely on it telling the truth.


The WiFi network has been running perfectly for about 18 months, with more iot devices than I've got now.
About 6 months ago I replaced about 20 WiFi smart bulbs with Matter over Thread bulbs, so there are less WiFi clients now - around 70.

The problem first started showing a few weeks ago, and it looked like it was just one of the two Tapo H100 hubs that was playing up - sensors that are linked to that hub going offline in Home Assistant randomly a few times a day.
I replaced that Hub.

Now, both Tapo hubs are going offline much more frequently, as well as the remaining WiFi smart bulbs, Shelly switches and EM devices, Tapo smart plugs, and the Ring chime and doorbell. Alexa devices also stop responding to voice commands (but don't reply with a "sorry, no internet" voice message).

After a minute or two, everything that's dropped offline automatically reconnects without me doing anything.


I've got it set up so that the Deco Mesh is running in Bridge mode, so the BT smart hub is acting as the router, doing the DHCP stuff.
I thought it was going to be a problem with the Decos, but is it possible that the BT router is the problem?
Even though the local WiFi is all handled by the Decos, would clients still need to "see" the the BT router because that's where they originally got their ip address from?

jimothyc

846 posts

113 months

Friday 21st August
quotequote all
Do you have the ethernet backhaul all connected to a managed switch? You might want to check the interface of that and double check line speeds etc. I had similar issues on my Ubiquiti system when after a power failure an ethernet connected access point decided to mesh with another access point and created a loop in the network which can cause all sorts of issues.

clockworks

Original Poster:

7,671 posts

174 months

Friday 21st August
quotequote all
jimothyc said:
Do you have the ethernet backhaul all connected to a managed switch? You might want to check the interface of that and double check line speeds etc. I had similar issues on my Ubiquiti system when after a power failure an ethernet connected access point decided to mesh with another access point and created a loop in the network which can cause all sorts of issues.
It's a fairly basic TP link poe switch. I can log into it and see some stuff. All I've done is enabled poe on 4 of the ports. Not sure if that counts as "managed" though?

I'll have a look.

The "master" Deco unit is connected directly to the BT router.

ARH

2,050 posts

268 months

Friday 21st August
quotequote all
Its not well reported but BT routers can only really cope with 32 devices connected reliably to each SSID, no idea if this is the case with your router or not, it certainly was with mine but it was many years ago now. This may not be an issue with newer routers, but I would not expect the issue to have gone away. The router will happily give the devices IP addresses, and reconnect to them when or if they disconnect, but sending data to them just get lost. With mine stuff would just randomly disconnect or reconnect, there was no pattern. All would work well for weeks then just go into meltdown, after fiddling about it came back, but would just fail again.

Nicetobenice

1,493 posts

7 months

Friday 21st August
quotequote all
Have you got the lights on a different network to the other devices?

That made a material difference to the performance of our network.

clockworks

Original Poster:

7,671 posts

174 months

Friday 21st August
quotequote all
Yes, all devices are on the same SSID, which is the Mesh network. Same SSID for 2.4 and 5ghz.

The router's onboard wifi uses a different SSID, which we use if there's a power cut - router and ONT are on UPS's.

Nicetobenice

1,493 posts

7 months

Friday 21st August
quotequote all
clockworks said:
Yes, all devices are on the same SSID, which is the Mesh network. Same SSID for 2.4 and 5ghz.

The router's onboard wifi uses a different SSID, which we use if there's a power cut - router and ONT are on UPS's.
It's a pain in the arse but see if you can create a different network for the smart lights.

The ones we have all use the 2.4ghz band so they are on a separate network along with other stuff like a washing machine and fridge (god knows why either of them have wifi)
Might also be worth switching the router wifi off until you need it.

clockworks

Original Poster:

7,671 posts

174 months

Friday 21st August
quotequote all
Nicetobenice said:
It's a pain in the arse but see if you can create a different network for the smart lights.

The ones we have all use the 2.4ghz band so they are on a separate network along with other stuff like a washing machine and fridge (god knows why either of them have wifi)
Might also be worth switching the router wifi off until you need it.
Swapping all the smart devices over to a separate SSID would indeed be a massive pain, as they would need to be factory reset and added back in via the manufacturer's apps, then re-added to Home Assistant. It would likely take a couple of days! Last resort I think.
The fact that it has all worked perfectly for well over a year makes me think that there's something more basic at the heart of the problem.


What I'm doing now is using another tablet running the Ping app, with it set to carry on running in the background so it stays running whatever the tablet does.
I switched off the tablet WiFi, moved it very close to one of the mesh points, and re-enabled the wifi.
This seems to have forced it to latch onto that particular wifi - confirmed in the Deco app after it has updated the map.
I then start the Ping app running, pinging that mesh device ip, and see if I get any packet losses.

I'll go round the mesh points one by one.

If that all passes, I'll do it again, but pinging the Master mesh device ip address.

Nicetobenice

1,493 posts

7 months

Friday 21st August
quotequote all
It could well be one device rather than the access point.

Might be worth recording if the situation is worse when certain devices are powered on.

I'd definitely put the router in modem mode as well, switching it on if you need it.

clockworks

Original Poster:

7,671 posts

174 months

Friday 21st August
quotequote all
Nicetobenice said:
It could well be one device rather than the access point.

Might be worth recording if the situation is worse when certain devices are powered on.

I'd definitely put the router in modem mode as well, switching it on if you need it.
Yeah, good point. One device could be flooding the network?


The problem with turning off the router's onboard wifi is we only need it if there's a power cut - stays working on a BT-supplied UPS.
With no power, I wouldn't be able to get into the router admin page to turn it on again, as we don't have a battery-powered laptop with an ethernet port that can be plugged directly into the router.

I can try it though, see if it's causing the problem.

Nicetobenice

1,493 posts

7 months

Friday 21st August
quotequote all
clockworks said:
Yeah, good point. One device could be flooding the network?


The problem with turning off the router's onboard wifi is we only need it if there's a power cut - stays working on a BT-supplied UPS.
With no power, I wouldn't be able to get into the router admin page to turn it on again, as we don't have a battery-powered laptop with an ethernet port that can be plugged directly into the router.

I can try it though, see if it's causing the problem.
We never really got to the bottom of what was causing our network to hang. It may have been one device or it could have been general congestion and conflicts.

Since I went through the process it's been very stable.