Myriota Connector – UplinkMessageProcessor Queue Output Binding

The myriota Azure IoT Hub Cloud Identity Translation Gateway uplink message handler Azure Storage Queue Trigger Function wasn’t processing “transient” vs. “permanent” failures well. Sometimes a “permanent” failure message would be retried multiple times by the function runtime before getting moved to the poison queue.

After some experimentation using an Azure Storage Queue Function Output binding to move messages to the poison queue looked like a reasonable approach. (Though, returning null to indicate the message should be removed from the queue was not obvious from the documentation)

[Function("UplinkMessageProcessor")]
[QueueOutput(queueName: "uplink-poison", Connection = "UplinkQueueStorage")]
public async Task<Models.UplinkPayloadQueueDto> UplinkMessageProcessor([QueueTrigger(queueName: "uplink", Connection = "UplinkQueueStorage")] Models.UplinkPayloadQueueDto payload, CancellationToken cancellationToken)
{
...
   // Process each packet in the payload. Myriota docs say only one packet per payload but just incase...
   foreach (Models.QueuePacket packet in payload.Data.Packets)
   {
      // Lookup the device client in the cache or create a new one
      Models.DeviceConnectionContext context;

      try
      {
         context = await _deviceConnectionCache.GetOrAddAsync(packet.TerminalId, cancellationToken);
      }
      catch (DeviceNotFoundException dnfex)
      {
         _logger.LogError(dnfex, "Uplink- PayloadId:{0} TerminalId:{1} terminal not found", payload.Id, packet.TerminalId);

         return payload;
      }
      catch (Exception ex) // Maybe just send to poison queue or figure if transient error?
      {
         _logger.LogError(ex, "Uplink- PayloadId:{0} TerminalId:{1} ", payload.Id, packet.TerminalId);

         throw;
      }
...
         // Proccessing successful, message can be deleted by QueueTrigger plumbing
         return null;
      }

After building and testing an Azure Storage Queue Function Output binding implementation I’m not certain that it is a good approach. The code is a bit “chunky” and I have had to implement more of the retry process logic.

Leave a comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.