AMQP/MQTT: Gateways may subscribe for commands to disabled devices

[calohmn]

Currently, when a gateway subscribes for commands of a specific device, there is no check whether that device is actually enabled.

See this example trace:

The command subscription is successful - only sending the notification event fails.

(When a northbound application actually tries to send a command to that device, that request will be rejected however - the device registration data is checked when trying to find the mapped gateway.)

[sophokles73]

While I tend to agree that the adapters should reject a subscription request if the device is disabled at the point in time of the subscription request, I wonder how we want to deal with a situation where the device is being disabled after the subscription has been created?

[sophokles73]

@calohmn any additional thoughts?

[calohmn]

I wonder how we want to deal with a situation where the device is being disabled after the subscription has been created?

If we implement a periodic command subscription renewal, done by the protocol adapters (to prevent stale commandHandlingAdapterInstance entries in the Device Connection service - see #1858), we could also check the device registration at these points in time.
When finding out about a disabled device, we could invoke the handler, suggested in #1859, to close the subscription (meaning closing the MQTT connection, for example).
The question is, if closing the subscription and thereby explicitly informing the device that it won't receive commands, is all that important. It's the same question also raised in #1859. I tend to think it's probably not that important.

[sophokles73]

@calohmn IMHO the least we should do is perform the check during subscription establishment.

[calohmn]

[...] I wonder how we want to deal with a situation where the device is being disabled after the subscription has been created?

The fix for #293 includes closing a MQTT device connection if the device got disabled. In case an MQTT gateway is connected and a device connected to that gateway gets disabled, any existing command subscriptions for that disabled device are left unchanged at the moment.
But I think that behaviour should be kept as is, since the MQTT adapter can not inform the connected gateway that certain command subscriptions have now been removed. If we would just remove the command subscription handler for such a device in the MQTT adapter, the MQTT gateway wouldn't know that it has to SUBSCRIBE again when the device gets enabled again at some later point.

IMHO the least we should do is perform the check during subscription establishment.

@sophokles73 yes, that check should be added.

[calohmn]

IMHO the least we should do is perform the check during subscription establishment.

Actually, that check has already been added in #2377.