Elixir, Fly, and Grafana Cloud

Published

If you have an elixir app running on Fly.io, sending your telemetry data to Grafana cloud is easy.

Lets start with Logs.

Logs

In order to ship logs to Grafana Cloud, we need to setup another application called logshipper. It is all detailed in the documentation, but the TLDR; is:

Logshipper Documentation

  • Create a new directory
    mkdir logshipper
  • Create a new app, but don’t launch yet
    fly launch --no-deploy --image ghcr.io/superfly/fly-log-shipper:latest`
  • Configure your org and access token
    fly secrets set ORG=personal
    fly secrets set ACCESS_TOKEN=$(fly auth token)
  • Configure your Loki credentials. These can be found in the Grafana Cloud Portal, Loki section. You may have to generate an API key, but all the data should be there for you to use here.
    fly secrets set LOKI_URL=
    fly secrets set LOKI_USERNAME=
    fly secrets set LOKI_PASSWORD=
  • Add this to the newly generated fly.toml file:
    [[services]]
      http_checks = []
      internal_port = 8686
  • Deploy
    flyctl deploy

Once deployed, this should start sending all the logs from all your fly apps in the configured organization to Grafana Cloud.

More configuration options for the logshipper app can be found in the github repo

Also, I’d recommend shipping your logs in JSON format using something similar to the logger_json library.

The logger_json library

Traces

Traces are fairly easy to send to Grafana Cloud Tempo, but we’ll need to start by adding the OpenTelemetry libraries. Depending on your needs, you may leave off some of these.

# ./mix.exs
defp deps do
  [
    ...
    {:opentelemetry_exporter, "~> 1.0"},
    {:opentelemetry, "~> 1.0"},
    {:opentelemetry_api, "~> 1.0"},
    {:opentelemetry_ecto, "~> 1.0"},
    {:opentelemetry_liveview, "~> 1.0.0-rc.4"},
    {:opentelemetry_phoenix, "~> 1.0"},
    {:opentelemetry_cowboy, "~> 0.2"}
  ]
end

Now we need to add some configuration to the config/runtime.exs file.

# ./config/runtime.exs
if config_env() == :prod do
  otel_auth = System.get_env("OTEL_AUTH") ||
    raise """
    OTEL_AUTH is a required variable
    """

  config :opentelemetry_exporter,
    otlp_protocol: :grpc,
    otlp_traces_endpoint: System.fetch_env!("OTLP_ENDPOINT"),
    otlp_headers: [{"Authorization", "Basic #{otel_auth}"}]
end

Next, setup the environment variables.

  • The value required for OTLP_ENDPOINT can be found on the Grafana Cloud Portal Tempo section, and will look something like: https://tempo-us-central1.grafana.net/tempo (your url may be different)
  • The value for OTEL_AUTH is a base64 encoded value of {username}:{api token} which can be easily obtained using:
    echo -n 'username:password' | base64`
    (replace username and password with the actual vaules)
  • And you’ll need the data source name (found on the same Grafana Cloud Portal page) which will be used to set the value of OTEL_RESOURCE_ATTRIBUTES

All of these values can be set with one command:

flyctl secrets set OTLP_ENDPOINT=https://your_endpoint OTEL_RESOURCE_ATTRIBUTES=your_datasource_name OTEL_AUTH=your_base64_encoded_string

After setting these values and deploying the application, traces should start showing in Grafana!

Metrics

For Metrics, I really like to use Prometheus and I find the easiest way to get started with Prometheus is using prom_ex. PromEx provides excellent documentation and is worth reading, but a quick guide to get it working:

Prometheus Overview

Documentation for prom_ex

  • Add :prom_ex, "~> 1.8" to your dependencies in mix.exs and run mix deps.get
  • Run the generator
    mix prom_ex.gen.config --datasource curl`
  • Add config in config.exs for the metrics server
    config :your_app, YourApp.PromEx,
        metrics_server: [
        port: System.get_env("PROM_PORT") || 9091,
        path: "/metrics",
        protocol: :http,
        pool_size: 5
    ]
  • Add YourApp.PromEx to your supervision tree in application.ex
    def start(_type, _args) do
        children = [
          YourAppWeb.Endpoint,
          # PromEx should be started after the Endpoint, to avoid unnecessary error messages
          YourApp.PromEx,
          ...
        ]
  • Lastly, uncomment any desired plugins in the generated YourApp.PromEx file.

With all of that, localhost:9091/metrics should be available when the app is running

Next, the metrics just need to be exposed so that Fly can scrape them. As documented by Fly, just add the following to your fly.toml file:

Fly documentation for metrics

[metrics]
port = 9091
path = "/metrics"

Finally, setup the Prometheus data source in Grafana Cloud with the following properties:

  • HTTP -> URL “https://api.fly.io/prometheus/<org-slug>/“ where org_slug is your org
  • Custom HTTP Headers -> + Add Header:
  • Header: Authorization, Value: Bearer <token> where token is the result of flyctl auth token

You should now see all of the fly metrics and prom_ex defined metrics in Grafana!

Wrap Up

I wrote this because I wanted to add observability to my Elixir/Phoenix apps that run on Fly and finding the information I needed was scattered throughout the docs. Hopefully others find this useful.

Happy Observing!