Generic Network problems using ESP-Arduino library 2.4.0

[aferraz]

Hi @arendst,
after upgrading to the release v5.12.0a, 2 days ago, I notice reachability problems on all my NodeMCU and Electrodragon board. In the log I notice continuous connection attempts to the mqtt broker and ping lost for all boards.
Following ad example:
Feb 13 18:28:34 esp14 esp14-1845 ESP-MQT: Attempting connection...
Feb 13 18:28:34 esp14 esp14-1845 ESP-MQT: Connected
Feb 13 18:28:34 esp14 esp14-1845 ESP-MQT: tele/esp14/LWT = Online (retained)
Feb 13 18:28:34 esp14 esp14-1845 ESP-MQT: cmnd/esp14/POWER =
Feb 13 18:29:04 esp14 esp14-1845 ESP-MQT: Attempting connection...
Feb 13 18:29:04 esp14 esp14-1845 ESP-MQT: Connected
Feb 13 18:29:04 esp14 esp14-1845 ESP-MQT: tele/esp14/LWT = Online (retained)
Feb 13 18:29:04 esp14 esp14-1845 ESP-MQT: cmnd/esp14/POWER =
Feb 13 18:29:44 esp14 esp14-1845 ESP-MQT: Attempting connection...
Feb 13 18:29:44 esp14 esp14-1845 ESP-MQT: Connected
Feb 13 18:29:44 esp14 esp14-1845 ESP-MQT: tele/esp14/LWT = Online (retained)
Feb 13 18:29:44 esp14 esp14-1845 ESP-MQT: cmnd/esp14/POWER =
Feb 13 18:30:14 esp14 esp14-1845 ESP-MQT: Attempting connection...
Feb 13 18:30:14 esp14 esp14-1845 ESP-MQT: Connected
Feb 13 18:30:14 esp14 esp14-1845 ESP-MQT: tele/esp14/LWT = Online (retained)

Some minutes ago I'm back to version 5.9.1i and the connection attempts have disappeared.

I have also updated all my POWs to version v5.12.0a but no problems in this case.
What do you think can cause the problems?

Thank you
Alessandro

[arendst]

I've seen this too but only on some devices and only with 2.4.0 library. Workaround is using the (2.3.0) pre-compiled release images or compile it yourself using the 2.3.0 library. I then was able to upload a 2.4.0 image which works well too.

I think the problem is with the SDK version used in 2.4.0 using part of flash to store wifi performance parameters which might get overwritten by my flash usage. I will investigate but it will take time as I have to create the issue and get flash dumps to pinpoint the issue.

The first step is to find out where the esp/arduino team has decided to store the SDK wifi performance parameters in flash as this is a user controlled area to made known to the SDK using a fixed callback routine.

[aferraz]

@arendst I tried it seems to work.
I leave this issue open for the benefit of others, if you think it's appropriate, you can close it.

Thank you

[aferraz]

@arendst after about 1 hour of uptime, with version 2.4, the problem stared again.
I have recompiled with 2.3 and upgraded all boards. At the moment everything seems OK

[arendst]

Investigation so far revealed that my initial thought is not the cause of the problem; flash is not compromised or corrupted leading to the TCP/IP problems. In fact the flash area used is FCxxx (or 4 blocks down from the end of flash) which I do not use.

I'm currently running a staged (2.4.0 + latest commits) version which seems to run fine but it is too early to make any conclusions yet. This version contains some changes regarding lwip2

UPDATE: Still no problems with 2.4.0 + latest commits. If downgrade to 2.4.0 it immediately fails. More info in Chrome Developer Tools; main page ajax times out almost all of the time. Ota uploading 2.4.0 + latest commits works right away as does 2.3.0.

I'll stop monitoring and blame 2.4.0

[jjlawren]

I've been having issues with frequent disconnections on a Basic using 5.12.0e with the 2.4.0 SDK. After seeing this case I've rolled the SDK back to 2.3.0 and I haven't had any disconnections in the first 2 hours of uptime.

@arendst, is there a straightforward way to test the latest 2.4.0 + commits version you referenced with PlatformIO?

Edit: I guess I should be watching here: platformio/platform-espressif8266#76

Edit 2: Still going strong with zero disconnections after 8 hours on 2.3.0. Would drop every 5-30 minutes on 2.4.0.

[bigbuglamer]

If i understand correctly then it's fixed in 2.4.1

esp8266/Arduino#4105

[aferraz]

@arendst any news about this ?
Thanks
Alessandro

[arendst]

Your guess is as good as mine. I suggest you keep an eye on the esp arduino issue list. I'll wait for 2.4.2 as then some problems will be solved and others maybe introduced sadly.

[aferraz]

Thanks @arendst . I will

[ascillato2]

New 2.4.2 esp8266 library has been released. Please, use 2.3.0 or the latest. Avoid 2.4.0