I have 29 Elsys ERS sensors distributed around a 6 story building logging environmental data. These are configured to use OTAA and send data every 10 mins. These all worked fine when sending data to a single TTN gateway.
I installed 3 x LorixOne gateways in this building, one on the ground, one on the 3rd floor and one on the 5th floor to ensure coverage. All LorixOne gateways are connected to separate 4G modems and configured to talk to the ttn-router-eu router.
I discovered that there was random packet loss when all three gateways were operational. The middle gateway on the third floor fell over on the 24th of June as I forgot to top up the 4G sim card with credit. After this, I have been getting great reception from all sensors in the building.
Does anybody have any idea why the use of multiple gateways may cause packet loss?
I’ve attached a photo below showing the before and after, for a selection of sensors.
Any chance the devices are doing an OTAA Join for every request?
If not: do the nodes use ADR? Then I wonder if the culprit gateway was working just fine at the start and somehow failing some time later, after ADR already had instructed (some of) the nodes to use a better data rate (lower SF) and/or transmission power. That would have resulted in a lower reach, possibly making that culprit gateway the only one that received their uplinks. If no downlinks are used while ADR is enabled, then every 64 messages (ADR_ACK_LIMIT, fixed in LoRaWAN 1.0.x) a node will include an ADRACKReq MAC command in an uplink, to which the server should respond within the next 32 uplinks (ADR_ACK_DELAY for EU868). So, if the nodes’ uplinks were only received by the culprit gateway then at some point the nodes would be missing the ADR downlink frame, and change their settings (lower their data rate, so: increase their SF) until their uplinks were received by the other gateways again. With one message every 10 minutes this process might take some time, though 6 days (if I got the screenshot right) seems quite long.
If caused and resolved by ADR then that would not be affected by the 4G credit running out; that would just be a coincidence then. So if ADR is enabled then maybe the first questions should be: how instant were the changes, when did you last see the culprit gateway listed in the metadata of the uplinks, and did you see the SF change?
Also wondering: did the culprit gateway use more data than the other two? And did the packet loss apply to all sensors, or (roughly) only to those close to the culprit gateway? Did TTN Console show anything special? And is the sensor data mostly received by one gateway, or by both?
