I have been thinking about different ways to scale the ChirpStack Gateway Bridge recently, and one such way I’ve seen mentioned is to have dedicated e2e GWB for certain Gateways, ensuring that their data ends up in certain topics, and are consumed by certain GWBs.
In theory, can anyone confirm if this would in turn require an individual DB per GWB instance, or would it still be safe to have all N GWBs pointing to the same DB instance?
The GWB does not have a database of it’s own. It is stateless. If you are referring to the MQTT broker, you can easily have multiple GWBs pointed at the same broker. So no you do not need multiple of anything else to have multiple GWBs, just spin up multiple instances and you should be good to go. This would be a similar architecture to if you had a gateway bridge installed on each of your gateways.
Although to be clear the data wouldn’t end up in different topics based on what GWB it uses. The MQTT topic is based off the region and the gateway EUI.
If you are curious about more high-availability / redundancy measures I made a write-up on my own implementation I am currently working on (although I have my GWBs installed on the gateways):