Configuration of OpenThread regarding backoffs

RE: Configuration of OpenThread regarding backoffs

AHaug — Fri, 01 Dec 2023 08:43:16 GMT

Happy to help!

I did some more digging and it looks like I might've referred to the wrong retry parameter config. I also found MLE_MAX_TRANSMISSION_COUNT in the Thread specification, which is set to 3 https://github.com/openthread/openthread/blob/main/src/core/thread/mle.hpp#L770

Nonetheless, if that also is 3 then the information is not wrong, but if you want to change it then the first parameter might not be the correct one

Other than that, let me know if anything else is left unanswered! If everything is fine, I'll mark the case as verified answer :)

Kind regards,
Andreas

RE: Configuration of OpenThread regarding backoffs

Simon_89 — Wed, 29 Nov 2023 15:50:30 GMT

Thanks a lot, this then answers everything!

RE: Configuration of OpenThread regarding backoffs

AHaug — Wed, 29 Nov 2023 13:59:58 GMT

Hi,

Yes, 3 retries are default if the pack is not ACKed. From nrf5_sdk_for_thread_and_zigbee_v4.2.0_af27f76\external\openthread\include\config\mac.h it is stated

* @def OPENTHREAD_CONFIG_MAC_DEFAULT_MAX_FRAME_RETRIES_DIRECT
 *
 * The default maximum number of retries allowed after a transmission failure for direct transmissions.
 *
 * Equivalent to macMaxFrameRetries, default value is 3.
 *
 */
#ifndef OPENTHREAD_CONFIG_MAC_DEFAULT_MAX_FRAME_RETRIES_DIRECT
#define OPENTHREAD_CONFIG_MAC_DEFAULT_MAX_FRAME_RETRIES_DIRECT 3
#endif

Kind regards,
Andreas

RE: Configuration of OpenThread regarding backoffs

Simon_89 — Wed, 29 Nov 2023 12:07:59 GMT

It seems that the backoff is visible in the parent request workflow.

The open question is still, if it is correct that we typically get the parent requests in triplets?

Thanks,

Simon

RE: Configuration of OpenThread regarding backoffs

AHaug — Wed, 29 Nov 2023 11:24:49 GMT

Aha, does this mean that your questions have been answered or are there anything left unclear regarding this topic?

Kind regards,
Andreas

RE: Configuration of OpenThread regarding backoffs

Daniel B — Wed, 29 Nov 2023 07:50:21 GMT

Turns out that the parent requests do exhibit an exponential backoff, the backoff is between groups of three (3) parent requests which get sent in short succession. As seen here

RE: Configuration of OpenThread regarding backoffs

AHaug — Tue, 28 Nov 2023 13:44:15 GMT

Hi,

I've heard with some of my colleagues regarding this and I have some more input to supply you

[quote user="Simon_89"]Yes, we were looking at the sniffs using Wireshark. To which workflows should the backoffs be applied?[/quote]

The backoff should be applied to the Parent request workflow you mention.

You should also be able to test the backoff by putting a device in SED and then turning off every router in the network that the device is connected to

Could you also supply the pcap and any necessary network keys for a sniffer trace where you see the issue present if the issue still persists?

Kind regards,
Andreas

RE: Configuration of OpenThread regarding backoffs

Simon_89 — Fri, 24 Nov 2023 07:00:42 GMT

Hi Andreas,

thanks for the reply! Yes, we were looking at the sniffs using Wireshark. To which workflows should the backoffs be applied?

- Discovery requests (when no commissioned network in memory) using the joiner module

- Parent requests (when commissioned network is not available anymore) after setting threadEnable()

- other possibilities?

We are using nRF2 SDK for Thread and Zigbee 4.2, so some older OpenThread. If I recall correctly, it's Thread 1.1

Best,

Simon

RE: Configuration of OpenThread regarding backoffs

AHaug — Mon, 20 Nov 2023 14:37:17 GMT

Hi,

Based in the definition of said config in https://github.com/openthread/openthread/blob/6e7ed5cb554876f83c5ae496ccfd4818228540a5/src/core/config/mle.h#L128 it looks like the backoff feature should be automatically included in the Openthread stack and I agree with you w.r.t what you think the configuration should do.

If I would guess as to why you don't see this backoff interval increase (up to a maximum of 20 minutes as defined by OPENTHREAD_CONFIG_MLE_ATTACH_BACKOFF_MAXIMUM_INTERVAL) it would be that you don't have any missing packs.

Did you also investigate a sniffer trace to check if you actually have any missing replies? If not, could you check and verify that?

Kind regards,
Andreas