Mesh network oscillating between 1 and 2 partitions

[fgervais]

I have 2 border routers on the 2 opposite side of my home.

Most of the time, they don't really see each other. They see the MLE advertisements but it's below OPENTHREAD_CONFIG_MLE_PARTITION_MERGE_MARGIN_MIN so they split in 2 partitions and everything is good.

However, I guess that the MLE link margin is just below OPENTHREAD_CONFIG_MLE_PARTITION_MERGE_MARGIN_MIN because at some point they feel like merging partitions, one of them becomes child, drops childs etc etc.

The devices associated with this BR loose network access for a little while.

Then short after they figure out that merging was a bad idea and the split again. Again, device stops responding for a little while.

This is only about 12 hours on one of the BRs:

counters mle
Role Disabled: 0
Role Detached: 43
Role Child: 22
Role Router: 0
Role Leader: 21
Attach Attempts: 21
Partition Id Changes: 43
Better Partition Attach Attempts: 330
Parent Changes: 0

I don't really understand because I feel like OPENTHREAD_CONFIG_MLE_PARTITION_MERGE_MARGIN_MIN is there for this exact use-case.

What could make my case special for it to fail?

Is my best solution to bump OPENTHREAD_CONFIG_MLE_PARTITION_MERGE_MARGIN_MIN to 20 and rebuild?

[jwhui]

Increasing OPENTHREAD_CONFIG_MLE_PARTITION_MERGE_MARGIN_MIN to increase the hysteresis range should help.

Separately, Thread-over-Infrastructure (TREL) allows BRs to incorporate infrastructure links (i.e. Wi-Fi or Ethernet) into the Thread topology and is intended to directly address this situation. It basically allows BRs to transport IEEE 802.15.4 frames over UDP/IP. If you like, you can try it out.

You can enable TREL in your build by setting the OTBR_TREL build option.

You can see if TREL is enabled in your build with the trel CLI command.

[fgervais]

Excellent, I will look at TREL in more details and give it a try.

As usual, thank you for taking the time to respond, much appreciated.