MLE questions

[ksathara]

Hi,
We have multiple SEDs and multiple FTDs, all having the same PANID and network key. Each SED has the IPV6 of its associated or paired FTD and send info to this IPV6. Now since there are multiple FTDs, there could be multiple hops being made in this transaction, which is fine. The issue we are facing is that after an SED or couple of SEDs working like this for week or months, all of a sudden this path seems broken. None of the FTDs change their locations. There may be some turned off and back on. We have verified the SEDs are sending the data continuously. What has broken is one/more hops in between FTDs to the destination.

1. When an SED connects to a FTD via MLE, does it evaluate the full path between it and its final destination? We think not as MLE has no routing info for application level packets.
2. Else the SED connects to whatever FTD it has access to, then who's responsibility is to deliver the message? Suppose the hop in the middle gets disconnected, the SED will not know about it. Then the last FTD has to find another path I guess. But in our case it seems not.

We are using Nordic SDK 2.3.0. I dont know which OpenThread version is being used. When I tried version command in CLI, I get
'OPENTHREAD/2060039bb-dirty; Zephyr; May 20 2024 14:53:50'

Cheers,
Kaushalya

[jwhui]

The commit you referenced is 2060039bb

When an SED connects to a FTD via MLE, does it evaluate the full path between it and its final destination? We think not as MLE has no routing info for application level packets.

SEDs only look to attach to a parent. Thread is a general purpose networking layer - it supports any-to-any communication and does not optimize a path to a given destination.

Else the SED connects to whatever FTD it has access to, then who's responsibility is to deliver the message? Suppose the hop in the middle gets disconnected, the SED will not know about it. Then the last FTD has to find another path I guess. But in our case it seems not.

Routers are responsible for maintaining paths to all other routers.

Are you able to provide packet traces and logs to provide more visibility?

[ksathara]

Thanks. Yes I can understand a router can be monetarily disconnected due to excessive radio noise, but why doesn't it heal? When I shutdown the other routers, all the sensors connected to the destined router and started sending back data. So that means at least in that instant, there was a path to the destination. Why didnt the parent router (one the SED connected to) couldn't/wouldn't find that path?

Cheers,
Kaushalya

[ksathara]

Anyone know a way to trace the real time routes of packets in openthread? I see there is mesh diagnostics, not sure how to use it for this purpose.

[jwhui]

Logs and packet captures are best to provide more visibility.

Without more visibility, it could be anything - including external RF conditions.

[ksathara]

You mean wireshark logs? The problem I have is my wireshark logs cannot decrypt some packets. How can you track in that scenario? I have added a packet trace. For ex. in the attached log I have a sensor - rloc 0xa017. Easily I can see it sends packets to 0xa000. But from that point onwards how can I trace?
Thanks,
Kaushalya
Apr-24-2.zip

[Irving-cl]

Could you provide the network key of the Thread network?

[ksathara]

Hi Li, Can we make this private? Unfortunately I cant share the networkkey in public domain. Cheers, Kaushalya

…

On Fri, May 24, 2024, 4:16 PM Li Cao ***@***.***> wrote: Could you provide the network key of the Thread network? — Reply to this email directly, view it on GitHub <#10277 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AE3WSE3KMGGU3CSM2WVSEBDZD3LNBAVCNFSM6AAAAABIAZ2G7KVHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM4TKNBTGEZTG> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[ksathara]

Thanks Jonathan, sorry about long list of questions. I couldn't find clear answers from OpenThread specification.

In that case, if a SED connects to a parent in a different partition, all its data intended for a recipient in the previous partition, will be lost. And there is no way to recover from this. Is my understanding correct?

If that is correct, then having partitions is a big problem, isnt it?

Cheers,
Kaushalya

[jwhui]

You should ensure your topology has enough router-capable devices to avoid partitions.

If your mesh is sparse enough to have partitions, there is not much redundancy and the benefits of mesh are not realized.

[ksathara]

Hi, Sorry for my delay as I was not well past week.

Is there a possibility that a SED can do a MLE attach and make sure it gets a 'new parent' each time? i.e. like the SED 'walk-through' each accessible parent in the SED's range until some condition is met.

[jwhui]

Such a capability would not be conformance to the Thread Specification. OpenThread does not implement such a capability.

[ksathara]

ok, thanks Jonathan. Much appreciated.