Troubleshooting Node Connectivity Issues in Thread Network

[dsyx]

I have two networks (A and B): A contains 15 nodes and 1 OTBR, B contains 24 nodes and 1 OTBR. All nodes joined the network using the following commands (network keys are different for A and B):

dataset networkkey 81ae2c2c17368d585dee71eaa8cf1e90

dataset commit active

ifconfig up

thread start

Then I pinged all nodes in the network in a loop: A worked fine for two weeks, and B worked fine for two months.

Recently I found that one of the nodes in A (called A1) was not accessible (ping failed). Checking the node status through A1's VCOM, I found that it became the leader and was not in A (it looked like A1 created a new network). Why did this happen? This is inconsistent with my expectation: A1 should be in A or not in any network. How should I solve this problem?

Thanks.

[jwhui]

A node typically becomes a Leader when it is not able to communicate with any neighboring devices. In many cases, this is due to a platform specific radio driver issue. Are you able to validate whether the device is sending and receiving IEEE 802.15.4 frames?

[dsyx]

How to ensure that device A1 is in network A or not in any network? If this is not possible, when A1 becomes the leader but cannot communicate with any device in A, A1 is useless (it does not know that it lives on an isolated island).

[jwhui]

In Thread, simply being configured by the Network A's Operational Dataset makes it a part of Network A. Because wireless connectivity is time-varying, it is possible for a device not to have radio connectivity to neighboring devices. In this case, the device itself will become a Leader. But as I mentioned previously, OpenThread has no way to tell the difference between lack of connectivity due to physical topology / environment vs. lack of connectivity due to bugs in radio driver implementation.

[dsyx]

In practice, it occasionally happens that a device becomes an island (such as A1 in the example). Is there any way to handle such cases? Are there any relevant documents or cases I can refer to?

Thanks.

[dsyx]

In the API reference I found the function otThreadSetLocalLeaderWeight. If I set OTBR to the highest weight and the other nodes to the lowest weight, is it feasible and reasonable to determine that the node is not in the original network when it becomes the leader?

[jwhui]

In Thread terminology, a network is simply the set of devices that are configured to the same Operational Dataset. A Thread network may be composed of multiple Thread partitions. Devices that are not able to connect via radio transmissions form their own partitions.

As I tried to say earlier, a device forms its own partition when it is not able to communicate with neighboring devices. This could be due to environment where radio transmissions cannot get through or could be due to bugs in the radio driver. I strongly suggest validating that the radio driver is operating correctly when you notice a device has become a Leader and formed its own Thread partition.