Incomplete/misleading "hono.downstream.*" metrics

The term “downstream” in Hono is usually associated with the direction from the device to the messaging network and from there to the Business Application. Therefore, my expectation for the hono.downstream.* protocol adapter metrics would be that these are about the messages sent from the protocol adapter to the messaging network.

However, that is not exactly the case. The message transfers covered by these metrics are:

• telemetry/event/command-response messages sent to the AMQP messaging network
• AMQP messages sent to the device registry (for retrieving tenant/device/credentials information)
• AMQP messages sent to the command router (for updating command subscription information)

Missing are telemetry/event/command-response messages sent to Kafka.

I think it would be better to have the AMQP messages sent to device registry and command router be covered by a different metric (not with “downstream” in the name).

After having added Kafka support, the documentation should also emphasize that hono.downstream.full (“The number of times a message should be sent, but could not because the sender was out of credit.”) only refers to AMQP message transfers.

[calohmn]

But it is not only about the delivery to the downstream messaging infrastructure but also includes the time it takes to actually process the message within the adapter, or isn’t it?

Right, the “delivery duration” term should only be applied to the “hono.amqp.*” metrics. So, it’s hono_telemetry_processing_duration and hono_amqp_delivery_duration then.

[sophokles73]

don’t forget to create a corresponding API change entry in the release notes …