High Availability Docs

[offizium-berndstorath]

I deploy your solution (which i contribute back to ofc) at a larger scale for some time.

But I have Chirpstack, Redis, Mqtt and Postgres on a single server. Which if down is very bad.

There is little to non actual examples and docs about HA. I know I can deploy some parts multiple times and load balance them. But how exactly? What would a good solution with 2 Servers look like?

I though about something like that. But I still have a single point of failure with Server1

flowchart LR
    subgraph Server1
        A[Gateway Bridge] --> |MQTT| Mosquitto
    end

    subgraph Server3
        Postgres1

        Chirpstack1
        Chirpstack1 --> |MQTT| Mosquitto

        Redis1

        HAProxy1
        HAProxy1 --> |SQL| Postgres1
        HAProxy1 --> |TCP| Redis1
        HAProxy1 <--> Chirpstack1
    end

    subgraph Server4
        Postgres2

        Chirpstack2
        Chirpstack2 --> |MQTT| Mosquitto

        Redis2

        HAProxy2
        HAProxy2 --> |SQL| Postgres2
        HAProxy2 --> |TCP| Redis2
        HAProxy2 <--> Chirpstack2
    end

    HAProxy1 <--> HAProxy2

    C[Load Balancer]
    Gateway --> |Semtech UDP| A
    C --> HAProxy1
    C --> HAProxy2
    User --> |HTTP| C

I am not 100% familiar with HA and Load Balancing so I don't know exactly if it would work like that

Maybe Chirpstack could be used with Kubernetes?

[brocaar]

There are multiple ways to set a HA ChirpStack instance. In general:

• You can scale ChirpStack horizontally (the MQTTv5 shared-subscription will take care of the uplink load-balancing)
• All ChirpStack instances must point to the same Redis and PostgreSQL
• Different strategies exist to create a HA setup of Redis, PostgreSQL and the MQTT broker (some brokers provide clustering support, others bridging, ...)
• Yes, the above could be achieved using Kubernetes

[offizium-berndstorath]

The point was to maybe add Docs or (concrete) Examples how HA / Failover could be implemented.