ADR Issue Dropping Devices?

We have a V4 LNS up and running using AU915_5 (Band 6). We have an issue with what looks like the ADR settings in the .toml file. The test devices are Dragino LHT65s and the Device Profile is set as V10.3 Rev A.

The problem we’re experiencing is this - The devices start off at 20minute uplinks, then the DR is changed from 2, to 1, to 0… and the uplinks start to disappear.
Where to start? The ADR margin was 10, changed to 12 and the problem is still there… have just changed it to 15… am I on the right track?

Thanks,
Mark

Please note that ChirpStack never lowers the data-rate. It looks more like the ADR backoff algorithm (see LoRaWAN Specifications for all the details). If this is the case, then it is probably triggered by packet-loss on the downlink path.

… this one was resolved in the interim by extending the device RX1DL and RX2DL times. This was the clue to look into the server’s response times.

In the end it was as simple as changing the AWS instance type from T1 to T3 - one of the differences between the two is the network response.

Problem solved.

Thanks for the response.