One Node, Multiple Gateways, too much "join session"

ChirpStack Network Server
#1

Hello There,
I’m using multiple gateway and i have some sensors in the “cover range” of both two gateway.
at the begining i see the RSSI pass from one to another and i said greaat it working well and all is combine to make a unique payload message. But today i see some problems

After few days of use. i get one sensor which have initiated over 700 “join” to the network. As if each time it switch gateway it start a new join session.

First impact, it totally flood the table device_join . could i delete all the useless field ?
second and more important problem : after move the sensor from one place to another this one try to reconnect with the error: “validate dev-nonce error”

Is it possible, there is a link between the fact that the sensor use to much join session and the dev-nonce error ???
some one know how to reset the dev-nonce?
i try to delete-recreate the device… without succes.

#2

I’ve experienced this due to firmware bugs on the device side. The dev-nonce error is because the device is attempting to join with a nonce that has already been used. This should never happen. Whether it was caused by the join storm, or was the cause of the join storm is for you to figure out.

You are correct, deleting the device does not clear out the nonces in the database. You will need to delete them manually from Postgres.

The gateways should be transparent to the device, so that seems less likely to be the immediate cause. Could it be that the device was joining before each uplink and thereby exhausted its nonces? I’ve seen that happen (again, device bug).

1 Like
#3

it seems you see right => bug at the node side.
with no reason after 48h both the two node with the nonce error have been reconnected without modifications at my side.

Just for understand better: could someone explain me, In the table “device_join” why i get different ID [pk]uuid for the same device?
some nodes have only one ID other have many and many IDs…
exemple here:

Capture d’écran 2021-02-08 151948
Capture d’écran 2021-02-08 1519481156×776 120 KB

because we talk about a “join storm” but i’m answering if its a “join BUG” or somethings different.
many thanks for your help

#4

I confirm it’s a “join session” ask from the node. i try to capture the lora frame when it append.
The result is clear and i try to explain the bug with this few steps:

  • consider the node run on the GW 1.
  • first the node found an other gateway near it , GW2. (materialized by two objects “RX info” in the payload with different GW IDs).
  • at the next payload it ask a join session.
  • the join session is validate by the LNS.

At the end, the glitch is, it got the new join session on the same gateway, GW 1. (materialized with a “Confirmed dataUp” with only one object “Rx_info” the GW1)

The GW2 is less interesting in RSSI and i dont undestand why the node try to ask a join session, as all seems to be good to it with the GW1.

i have screen shot the lora-frame from the AS, for help if it needed.

this, dont explain the “dev-nonce error” occur during few hours (roughtly 48hours) and re-establish “normaly” without config modifications. but this is a other issues