Downlink issues after migrating from Milesight UG65 embedded NS to ChirpStack v4 (TOO_LATE / DR not found)

Question

Hello,

I have a number of Milesight UG65 gateways, each originally using their own embedded network server. Recently, one of the master gateways for an application failed, so I decided to migrate everything to the cloud using ChirpStack v4 in Docker. I am currently performing tests before doing the full migration in a couple of weeks.

Because I have applications with nearly 150 flow meters deployed in the field, I want to avoid forcing all devices to rejoin the network (which would require physically resetting each flow meter).
So, I exported the SQL databases from the UG65 LoRaWAN NS, and also extracted the Redis .bin device-session files. Using this data, I recreated all applications, device profiles and devices in ChirpStack v4 through the API — including their activations. Since the Milesight NS is based on ChirpStack v3, the migration is not straightforward.

Uplinks work fine with my test devices — I receive all packets normally.
However, downlinks are the issue.

In ChirpStack → Device → Events, I consistently see:

DOWNLINK_GATEWAY: TOO_LATE

I also tried forcing a rejoin (because the original join was performed on the UG65 embedded NS). After the rejoin, I briefly saw correct ACKs as downlinks, and the device did receive downlinks from the queue — but only immediately after the join.

While searching the forum, I found this post:
https://forum.chirpstack.io/t/ack-not-being-seen-by-milesight-ug67-gateway/23678

which mentions that the “ChirpStack v4” option in Milesight gateways is outdated. Based on that, I tried Semtech UDP, but I saw the same issue. I also increased the rx1_delay, as commonly suggested, but it didn’t solve the problem.

Finally, I tried Basic Station, since I prefer not using UDP. With Basic Station, uplinks are received correctly, but no downlinks appear on the gateway, and in the gateway-bridge logs I see:

send downlink frame error:
downlink frame from proto error: get data-rate index error:
lorawan/band: data-rate not found

What could be happening?

I am using the chirpstack-gateway-bridge-basicstation-us915_0.toml file from the GitHub repository.
My device profile is configured for US915, channels 0–7 + 64.
In ChirpStack, the gateway also appears correctly under Region ID = us915_0.
I tried this migration with both OTAA and ABP devices.
I am attaching some screenshots.

Any guidance would be greatly appreciated.

Thanks!

Captura de pantalla 2025-11-28 105739.png

Orne Brocaar · Answer

DOWNLINK_GATEWAY: TOO_LATE

This usually indicates that the downlink payload arrived too late at the gateway, and the gateway is unable to send it at the requested timestamp.

In most cases, this is caused by a high network latency between the gateway and ChirpStack. By default in Class-A, the RX1 delay is 1 second. This can be controlled by changing the rx1_delay in the region_XXXXX.toml configuration (e.g. to rx1_delay=3). After changing this value (and restarting ChirpStack), ChirpStack will send a mac-command to the device, requesting to update the RX1 Delay configuration. Until the device acks this request, ChirpStack will continue to use a delay of 1 sec.

In most cases, making the aboove change should solve the issue once all devices have adapted to the new config.

Julia · Answer

Ok, I understand, thanks again. But as I mentioned before, I’m avoiding re-joining the network to prevent having to restart more than 200 devices in the field. Also, I changed the rx1_delay from the ChirpStack web interface under the device profile options.
This problem with rx1_delay might be related to the fact that I’m using the ChirpStack v4 option on the Milesight gateway, which I’ve read is outdated. What if I use Basic Station instead?
Why am I receiving the downlink frame error:
proto error: get data-rate index error: lorawan/band: data-rate not found
even though my band settings match, as shown in the screenshot above?

Downlink issues after migrating from Milesight UG65 embedded NS to ChirpStack v4 (TOO_LATE / DR not found)

2 Answers