[BUG] OTA Job notification isn't received

[Fienberber]

I'm migrating from to the esp-aws-iot [202406.01-LTS] from [202210.01-LTS].
The previous integration was working properly so I know this is not an AWS issue.

Using the OTA example, I can start OTA's using polling. Using the requestJobDocumentHandler(); by sending an event at OtaAgentEventRequestJobDocument.

But previously I was subscribed to $aws/things/+/jobs/#, which notified when OTA where available.
I don't want to use polling to start an OTA. But I don't achieve subscription to the jobs topic.

I think I'm missing a critical part of the example. Because it seems like it doesn't subscribe to the $aws/things/+/jobs/# topic.

[aggarg]

I think I'm missing a critical part of the example. Because it seems like it doesn't subscribe to the $aws/things/+/jobs/# topic.

Are you saying that your subscribe calls fails? Or the subscribe succeeds and you do not get a notification?

[Fienberber]

Thank you for your response,

Actually I thought I was subscribed but since I don't get the notification I tried to add the subscription manually.
At this moment I get disconnected from the broker.
In the example ota_over_mqtt_demo I don't see where the subscription take place.
I see the request during the initialization, but afterward I don't see any pulling either.

That's why I think I'm missing something.

[AniruddhaKanhere]

@Fienberber, the demo doesn't actively subscribe to the MQTT topic. See the jobs docs on AWS. Clients are sent the messages from the broker even when they are not subscribed to it.

To that end, the demos 'inject' an OTA handler which handles unsolicited publishes. Thus, any topics not handled by coreMQTT-Agent (or manager) will be passed on to the OTA handler which verifies whether the job topic matches with what is expected and processes it.

You should not need to process the OTA events yourself. The OTA task should do that on its own by calling processOTAEvents. Can you elaborate on what is happening? Is the task not starting? Are you using the unmodified version of the demo?

Thanks

[Fienberber]

Hi @AniruddhaKanhere thank you for you response.
I use a modified version of the demo in order to comply with infrastructure. However I get the OTA to work while polling for new job. I could have missed something when modifying the demo...

I enabled every level of logs for MQTT in the menuconfig. But I see no publish when I create a remote job on the AWS control panel. The OTA handler doesn't get anything either.
That's why I tried to do I myself, in order to see if the remote server actually publishes something.

The task seem to running since it works properly at startup. Or am I mistaken ?

[aggarg]

Can you try to add a log here to dump pxPublishInfo->pTopicName and pxPublishInfo->pPayload. This will confirm if the OTA notification is forwarded to the OTA handler.

[fjuliofontes]

Hi @Fienberber ,
Did you find the solution? I'm also suffering from the same issue. Only when calling requestJobDocumentHandler() the OTA starts.

[AniruddhaKanhere]

@fjuliofontes why do you need to call the requestJobDocumentHandler function manually? These lines should take care of the same.

Can you please add a breakpoint there and check whether that line is hit and the step through to figure out why isn't that message sent/processed?

[AniruddhaKanhere]

Also, can you try creating the job before running the code on the device?
I suggest, create the job on the console - wait for 30s, and then run the code on the device. That will verify whether there is some timing issue in the code in the repository.

@Fienberber and @fjuliofontes can you both please try the above?

[fjuliofontes]

@fjuliofontes why do you need to call the requestJobDocumentHandler function manually? These lines should take care of the same.

Can you please add a breakpoint there and check whether that line is hit and the step through to figure out why isn't that message sent/processed?

I'm not actually calling, it's during bootup that the function get's called. I was just pointing that it's the only way for the job to start, rebooting.

[fjuliofontes]

Also, can you try creating the job before running the code on the device? I suggest, create the job on the console - wait for 30s, and then run the code on the device. That will verify whether there is some timing issue in the code in the repository.

@Fienberber and @fjuliofontes can you both please try the above?

Thank you, as I described above the OTA will start at bootup if the job is already created.

By the way, the compiler was giving a warning about a falling-through on this line and I've added a break there.

Best regards,

[Fienberber]

Hello @AniruddhaKanhere thank you for your response.
Sorry for the delay, I still have the issue. I added a task that pulls jobs in the meantime.

Also, can you try creating the job before running the code on the device? I suggest, create the job on the console - wait for 30s, and then run the code on the device. That will verify whether there is some timing issue in the code in the repository.

@Fienberber and @fjuliofontes can you both please try the above?

As @fjuliofontes said, if the Job was created before the software start, it will work. At init the OTA is polled once. The issue happen when the software is running and a job is send from AWS.

Can you try to add a log here to dump pxPublishInfo->pTopicName and pxPublishInfo->pPayload. This will confirm if the OTA notification is forwarded to the OTA handler.

I've added the log, I've already done it, and didn't saw anything anormal.

Software startup
E (7155) ota_over_mqtt: vOTAProcessMessage Topis "$aws/things/DEVICE-ID/jobs/start-next/accepted", for Payload "{"clientToken":"test","timestamp":1729752877}"

• OTA Job creation
• Nothing even ater 10min
• Device REBOOT

E (7176) ota_over_mqtt: vOTAProcessMessage Topis "$aws/things/DEVICE-ID/jobs/start-next/accepted", for Payload "{"clientToken":"test","timestamp":1729754684,"execution":{"jobId":"AFR_OTA-****","status":"IN_PROGRESS","queuedAt":1729753816,"startedAt":1729754684,"lastUpdatedAt":1729754684,"versionNumber":2,"executionNumber":1,"jobDocument":{"afr_ota":{"protocols":["MQTT"],"streamname":"AFR_OTA-****","files":[{"filepath":"/device/updates","filesize":1367568,"fileid":0,"certfile":"/certificates/authcert.pem","sig-sha256-ecdsa":"MEQ...0Q=="}]}}}}"
E (14776) ota_over_mqtt: vOTAProcessMessage Topis "$aws/things/DEVICE-ID/streams/AFR_OTA-****/data/json", for Payload "{"f":0,"i":0,"l":4096,"p":"6QU...RQ=="}"
E (24727) ota_over_mqtt: vOTAProcessMessage Topis "$aws/things/DEVICE-ID/jobs/start-next/accepted", for Payload "{"clientToken":"test","timestamp":1729755015,"execution":{"jobId":"AFR_OTA-****","status":"IN_PROGRESS","statusDetails":{"self_test":"ready"},"queuedAt":1729753816,"startedAt":1729754684,"lastUpdatedAt":1729754989,"versionNumber":3,"executionNumber":1,"jobDocument":{"afr_ota":{"protocols":["MQTT"],"streamname":"AFR_OTA-****","files":[{"filepath":"/device/updates","filesize":1367568,"fileid":0,"certfile":"/certificates/authcert.pem","sig-sha256-ecdsa":"MEQ...0Q=="}]}}}}"
E (25737) ota_over_mqtt: vOTAProcessMessage Topis "$aws/things/DEVICE-ID/jobs/AFR_OTA-****/update/accepted", for Payload "{"timestamp":1729755016}"
E (30777) ota_over_mqtt: vOTAProcessMessage Topis "$aws/things/DEVICE-ID/jobs/start-next/accepted", for Payload "{"clientToken":"test","timestamp":1729755021}"

I can try other things if necessary.

[AniruddhaKanhere]

@Fienberber, thank you for that debug log.

Can you try to add the following line here:

    switch( recvEventId )
    {
+       case OtaAgentEventStart:
        case OtaAgentEventRequestJobDocument:
            ESP_LOGI( TAG, "Request Job Document event Received \n" );

            requestJobDocumentHandler();
            otaAgentState = OtaAgentStateRequestingJob;
            break;

I think that would fix the issue as the OTA task calls the requestJobDocumentHandler repeatedly.

[Fienberber]

@AniruddhaKanhere thank you for your response.

Unfortunately, it doesn't change anything. The processOTAEvents ins't even called after the Job creation...
I think the issue comes from the topic subscription.

[AniruddhaKanhere]

@Fienberber, I am not sure why doesn't these lines do not call processOTAEvents.

Can you please put a breakpoint there? The prvOTADemoTask should call it repeatedly.

As far as topic subscription is considered, the Jobs topic is a special topic to which IoT core publishes even without subscription as I highlighted above here.

Thanks for your patience.

[Fienberber]

@AniruddhaKanhere tank you for your response.

I've put a break point, actually it loops only 3 times, It it waiting for an event here.

The two previous events are : OtaAgentEventRequestJobDocument and then OtaAgentEventReceivedJobDocument as expected when the firmware starts.

But then it is just waiting for an event it never gets...

I don't know if it's expected but when OtaAgentEventReceivedJobDocument is received I run into this case (OtaPalJobDocFileCreateFailed).
The message received is on topic **/jobs/start-next/accepted with the content {"clientToken":"test","timestamp":1730280274}