Request size and concurrent job issues when using JobClient (.NET SDK)

Question

I have an Azure IoTHub (S2) environment with 4000+ devices, to which I'm trying to send the same TwinPatch to update them.
I am hitting 2 different issues when trying to achieve this update.

First
To achieve this update, I am using the Azure IoT SDK JobClient (.NET), using the following code:

JobResponse result = await this._jobClient.ScheduleTwinUpdateAsync(
                        Guid.NewGuid().ToString("D"),
                        $"DeviceId IN [{String.Join(',', deviceIds.Select(d => $"'{d}'"))}]",
                        twinPatch,
                        DateTime.UtcNow,
                        172800)

This results in the following error:

Precondition failed: {"Message":"ErrorCode:PreconditionFailed;The etag in the request does not match the etag of the existing resource, as per RFC7232, please check and try again","ExceptionMessage":"Tracking ID:26317b31a7c54068bf72889047952787-G:3-TimeStamp:11/18/2020 10:55:20"}

After some trial and error, I found that reducing the number of devices in the request to around 250 will make it succeed again. I can't seem to find any documentation on a limit within this query. Is there any limit, and if so, where is it documented?

Second
Subsequently, when splitting this job into multiple jobs, I am running into the concurrent job limit of 5 for an S2 instance (updating 4000 devices in batches of 250 takes 16 jobs). I tried scheduling every next job further in the future, and shortening the max runtime (so they don't overlap), but the 5 concurrent job limit seems to include Queued jobs. Which makes it completely impossible to use this as a primary scheduling mechanism, unless I'm missing something...

Answer 1

Hello @Frank Hoogmans ,

Below is the response from Microsoft's Product team, I hope this helps with your initial query.

1. It will update 10 per seconds and next will be subjects for https://learn.microsoft.com/en-us/azure/iot-hub/iot-hub-devguide-quotas-throttling#traffic-shaping and throttling.
2. However, it'll keep retrying throttled requests within specified job execution window. Hence, if windows is short there're chances it'll mark device job as throttled.
3. There's a separate bucket for jobs so it does not interfere with direct twin updates.
If you want to avoid throttling we recommend using https://learn.microsoft.com/en-us/azure/iot-hub/iot-hub-automatic-device-management