@David-TP

Thank you for your feedback and action plan.

1. Latest Deco version installed 
2. DNS settings change to primary 8.8.8.8, backup to 8.8.4.4
3. "Fast roaming and beamforming disabled.

After these changes, I've rebooted the whole network and deleted the devices listed as offline in the DECO App.

After about 1 hour there are 30x devices listed as online and 4x devices listed as offline, the offline devices change sometimes +/- 2x devices.

I did a scan from my PC which is connected via LAN and "Advanced IP Scanner" found 45x devices including the 5x Decos X50, this means there are 40x devices which have valid IP address and responded to a PING.

For example, I have an HP printer connected via WiFi and I can ping it from my PC via LAN and I can open the web interface on my mobile phone, but in the Deco app it's listed as an offline device, this also happens with devices connected via LAN, they are either not listed or change to offline devices.

The communication with LAN and WiFi devices is working and they are connected via MQTT or MODBUS and it would create a log entry if they lost the connection to LAN or WiFi but some of them are listed as offline devices from time to time or are event not listed in the DECO App.

This weekend I'm going to redesign the whole DECO network and change it from a WLAN router to an AP and use the ISP modem as a DHCP server, in the beginning I just used the ISP modem and two routers from different manufacturers and the ISP modem always showed the right number of devices connected.

Best Regards,

Stefan

  @SteWa I have the same behaviour with my Deco's. I have three m9 plus devices and recently bought a new main (deco x50 outdoor). I hoped this would solve the problem but some of the esp32 devices still disappear but I can ping them by their network name and ip address. 

 @David-TP

I had to enable "beamforming", the WiFi connection for the ESP8266 / ESP32 devices was unstable, each device make an automatic reboot when it lost the WiFi connection and this create an entry in the log of the MQTT broker over night I had more than 80x entries from around 10x devices.

The WiFi reboot is a setting that can be disabled, but I have found that it is necessary, I have also set a priority DECO for the ESP devices, otherwise after a reboot it will connect to the next DECO that is in the WiFi list of the device and this is not the one with the best signal, after a reboot it would connect again to the one with the best signal, now it is locked to a DECO.

Best Regards,

Stefan

 @David-TP

To make the behaviour more interesting, I've connected the 3x DECO X50 Backhauln <--> Switch TL-SG608E <--> TL-SG1016DE <--> TL-SG105E.

These SG608E & SG1016DE LAN switches are new, all LAN switches are TP-Link.

TL-SG608E --> 3x DECOS (OG, UG, Keller) and  Keller <-- WLAN --> Keller Zubau.

TL-SG1016DE --> Several LAN Devices 100 Mbit, 1000 Mbit LAN Devices)

TL-SG105E --> 1x DECO (Garage and LAN Devices)

My phone is connected to OG, now I move to UG, turn WiFi off/on on my Mobile Phone.

When this is done one or two IOT devices connected to UG reported a MQTT timeout but didn't lose the WiFi connection, it looks like every time there are changes done I/Os are on hold for ~500ms, maybe this is a known behaviour in MESH networks.

Anyway, I'm going to change the whole network from wifi router to AP, hopefully that will fix all the problems, if not I'll throw out the DECOS and go back to my NONE MESH setup.

Best Regards,

Stefan

  @BTJansen 

Today I have deleted the DECOs and added them again but as AP, Smart DHCP disabled, Fast Roaming and Beam forming enabled after rebooting the whole network, the Espressif are listed as online devices and do not go offline after almost 3 hours but all have the same name and no IP address in the DECO app.

Not all the IP devices are listed on the ISP Modem, which is now the DHCP server using 192.168.1.150-254 as the DHCP, IP below 192.168.1.150 are 20x static IP addresses.

The Espressif use a static IP address or DHCP, the Vendor is Sonoff, Shelly, Shelly PM, D1mini,ESP32-C3  with original firmware, Tasmota, Espeasy and the ESPs are ESP32, ESP8266.

On the ISP modem (DN9245W-10), 32x IP devices are listed as online, no offline devices.

  Listed are: Some with static IP addresses, DHCP and DHCP reserved.

DECO App (latest version, latest firmware on the X50) 29x IP Devices are listed as online, no offline devices

  Listed are: DHCP, and Espressif with static IP (without IP) and none of the DHCP reserved are listed  

Advanced IP Scanner, scanned 192.168.1.1 - 254 found 49x IP Devices

  Listed are: All IP devices

I have an ESP32-C3 which shows with RGB LED, status for booting, WLAN connected/failed, MQTT connected/disconnect.

When the DECOs were running in router mode the Boot to MQTT connection took about 2-3s, now it's so fast I can't see the change, it goes straight from booting to MQTT connected which is about 1s.

I will have to monitor the stability of the MQTT connection, but so far it looks more stable, with no timeouts reported in 3 hours.

Hello All,

I've finished my testing and analysis,

BUG: The ISP modem (DN9245W-10) should not be used as DHCP server, it has a bug and disable the new IP address assignment after 1-48h, reboot is required to fix this.

BUG: A DHCP server other than the one used on the DECO shows the ESP devices with an IP address while the DECO DHCP doesn't, it moves them to disconnected devices while the ping or http connection works.

ESP behavior, I've replaced an ESP8266 device which was reporting a higher number of timeouts within 24h in IOBroker for the MQTT ping with an ESP32 and now I have no timeouts and the whole 2.4 Ghz WiFi is more stable.

Note: I use a script that counts the MQTT timeouts / disconnects and every minute a ping is sent, all errors within 24h are summed up, every 24h a report is created and the counters are set to 0, this helps to identify failed devices or connection problems.

Hope this information's are useful,

BR, Stefan

 @BTJansen 

If possible, could you do the following test and let me know if you get the same results,

Open 3x CMD prompt in Windows (WLAN) and ping the following IP addresses with -t

- CMD1, [IP of an ESP1 device] -t

- CMD2, [IP of an ESP2 device] -t

- CMD3, [IP of a WLAN device, not an ESP 8266 or ESP32] -t

Let the ping run for a while, in my case up to 50ms (WLAN to WLAN) response time is reported.

Now check if one of the ESP devices is listed as a connected device in the DECO app, if so, reserve the IP address on the DECO for this device and the following happens in my network,

All ESP devices will be disconnected/connected, DHCP and fix IP, and the ping will report timeouts and response times up to 2900ms, this will take about 5min and it could happen that one or the other ESP device is disconnected for more than 5min.

In my case, the ESP rebooted always because it couldn't join the network, it tried but the DECO didn't give it an IP address.

I've seen similar behaviour in the past, it seems that the MAC address is still listed as connected and the DECO didn't allow the ESP to rejoin with the same MAC address.

Maybe that's why when I used my ISP modem as a DHCP server, the DECOS listed the Espressif as connected with a MAC address but no IP address.

In the next week I'll connect a old Router and use only the DHCP server of it and check the DECO behavior 

BR, Stefan
 

@David-TP

I have performed a PING from a Windows PC connected via LAN to an ESP32 WLAN device with a fixed IP address, each test performed 48 hours after a reboot of the DECOs.

The WLAN device (ESP32) was always connected to the same DECO WLAN and rebooted 5 minutes before the PING test.

A: The DECO network was running as a WLAN router with a DCHP server for 104x addresses ...150-254.

B: The DECO network was used as an Access point with Linux ICS DHCP on a Raspberry also for 104x addresses. 

The PING response in ms are much higher in A

A: (WLAN Router) The MQTT broker reported MQTT PING timeouts for the ESP devices, about 200x timeouts in 48h for around 10 devices.

The ESP devices are recognized as offline devices in the DECO APP, even if the PING works.

B: (Access Point) 1x timeout reported by the MQTT broker within 48h

The ESP devices are recognised in the DECO APP but without IP, PING works

The ESPressif are identified as Online devices but no IP is displayed, the device has a static IP and PING works.

The ESPressif devices with DHCP enabled are identified as Online devices and the IP address is visible.

Overall, it looks like the Access Point setup is more stable, as it's not possible to test the WLAN router setup without DHCP, as the DHCP server can't be disabled in this setup. I had to use the Linux ICS DHCP server with the Access Point Setup because the one on the ISP modem had a bug and stopped assigning new DHCP addresses between 1-48h.

Now everything works like a charm, no timeouts, no disconnects. 

Maybe this information is useful for others,

BR, Stefan