A Prometheus Exporter for Rivian

alt text Good to know: Unofficial Rivian Prometheus Exporter

Prometheus provides custom collectors to generate metrics to include in Observability platforms, you can write the exporters in supported client libraries.

I have written the Rivian Prometheus Exporter using the python client. You can deploy it using the Kubernetes resources below as well.

In the exporter, we have only one gauge (battery %) and couple of other counter metrics exposed.

Steps

Code Up Rivian Exporter in Python for Prometheus.
Dockerize the exporter, build container.
Deploy on Kubernetes
Scrape The Whip Metrics
Inspect the Data Outcome

Exporter

/src/rivian_exporter.py

import time
from prometheus_client.core import GaugeMetricFamily, REGISTRY, CounterMetricFamily
from prometheus_client import start_http_server
import rivian_api as rivian
import os
import json
import random

class RivianExporter(object):
    def __init__(self):
        self.rivian = rivian.Rivian()
        response = self.rivian.login(
            os.environ['RIVIAN_USERNAME'],
            os.environ['RIVIAN_PASSWORD']
        )
        # owner info, grab rivian vehicleid
        owner = self.rivian.get_user_information()
        self.rivianid = owner['data']['currentUser']['vehicles'][0]['id']
        print(f'Rivian: {self.rivianid}')

    def collect(self):

        # status info
        whipstatus = self.rivian.get_vehicle_state(self.rivianid)

        # battery level - batteryLevel
        # distance to empty - distanceToEmpty
        # gear status - gearStatus

        batterylevel = whipstatus['data']['vehicleState']['batteryLevel']['value']
        distancetoempty = whipstatus['data']['vehicleState']['distanceToEmpty']['value']
        gearstatus = whipstatus['data']['vehicleState']['gearStatus']['value']

        # Metric Translations
        if gearstatus == 'park':
            gearstatus = 0
        else:
            gearstatus = 1


        a = GaugeMetricFamily("rivian_battery_level", "% of Battery left", labels=['whip'])
        a.add_metric([self.rivianid], batterylevel)
        yield a

        b = CounterMetricFamily("rivian_battery_distance_empty", 'Miles Left', labels=['whip'])
        b.add_metric([self.rivianid], distancetoempty)
        yield b

        c = CounterMetricFamily("rivian_gear_status", '0=park, otherwise rolling...', labels=['whip'])
        c.add_metric([self.rivianid], gearstatus)
        yield c


if __name__ == '__main__':
    start_http_server(8000)
    REGISTRY.register(RivianExporter())
    while True:
        REGISTRY.collect()
        # lets not piss off the Site Reliability Teams at Rivian
        time.sleep(90)

I am exposing the static gauge and counter metrics. But you can just modify it with your logic from another system.

In second step we will build our Docker container.

Container

Container blurb.

FROM python:3.8

ADD src /src

RUN pip install prometheus_client
RUN pip install plotly
RUN pip install polyline
RUN pip install python-dateutil
RUN pip install python-dotenv
RUN pip install requests
RUN pip install geopy

WORKDIR /src


ENV PYTHONPATH '/src/'
ENV RIVIAN_PASSWORD 'secret'
ENV RIVIAN_USERNAME 'k8s'

CMD ["python" , "/src/rivian_exporter.py"]

Now build it, and push.

docker build -t rivian-prometheus-exporter .
docker image tag rivian-prometheus-exporter sween/rivian-prometheus-exporter:latest
docker push sween/rivian-prometheus-exporter:latest

Now that we have a container pushed, we can use it in our deployment.

Kubernetes

In the last third step you can deploy the collector on Kubernetes.\

/deploy/deployment.yaml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: rivian-prometheus-exporter-deployment
  labels:
    app: rivian-prometheus-exporter
spec:
  replicas: 1
  selector:
    matchLabels:
      app: rivian-prometheus-exporter
  template:
    metadata:
      labels:
        app: rivian-prometheus-exporter
    spec:
      containers:
      - name: rivian-prometheus-exporter
        image: sween/rivian-prometheus-exporter:latest
        ports:
        - containerPort: 5000
        env:
        - name: RIVIAN_PASSWORD
          valueFrom:
            secretKeyRef:
              name: rivian-user-pass
              key: rivian_password
        - name: RIVIAN_USERNAME
          valueFrom:
            secretKeyRef:
              name: rivian-user-pass
              key: rivian_username

Next is the service, I use MetalLB as it is my favorite thing ever basically, but expose the socket any which way you fancy.

/deploy/service.yaml

apiVersion: v1
kind: Service
metadata:
  name: rivian-prometheus-exporter-service
spec:
  selector:
    app: rivian-prometheus-exporter
  type: LoadBalancer # NodePort or whatever
  ports:
  - protocol: TCP
    port: 5000
    targetPort: 8000

Create a namespace and add your Rivian Credentials as a secret:

kubectl create ns rivian
kubectl create secret generic rivian-user-pass -n rivian \
    --from-literal=rivian_username='ron.sweeney+api@hotmale.com' \
    --from-literal=rivian_password='12345' # same as your luggage

Apply the Deployment and LoadBalancer (or NodePort)

kubectl apply -f deploy/* -n rivian

If everything worked out, we should see a pair of fonzies running on our cluster for the deployment and the MetalLB LoadBalancer Service.

Data Inspection

Now hit the MetalLB Load balancer on port 5000 and bask in the glory of the exported metrics.

Notice the sweet name spacing with my Vehicle Id

Though, awesome, most arent impressed by metrics endpoints, but get set to get even more unimpressed looking at a round trip of errands using prometheus to explore the data.

You need to define a simple prometheus.yml

# Sample config for Prometheus.
global:
  scrape_interval:     30s # By default, scrape targets every 15 seconds.
  evaluation_interval: 30s # By default, scrape targets every 15 seconds.
  # Attach these labels to any time series or alerts when communicating with
  # external systems (federation, remote storage, Alertmanager).
  external_labels:
      monitor: 'deezwatts'
# A scrape configuration containing exactly one endpoint to scrape:
# Here it's Prometheus itself.
scrape_configs:
  # The job name is added as a label `job=<job_name>` to any timeseries scraped from this config.
  - job_name: 'deezwatts'
    # Override the global default and scrape targets from this job every 5 seconds.
    scrape_interval: 30s
    scrape_timeout: 30s
    static_configs:
            - targets: ['192.168.1.92:5000']
    metrics_path: /

Then run it and hit http://localhost:9090 to explore the data time series in Prometheus.

docker run \
    -p 9090:9090 \
    -v $PWD/prom/prometheus.yml:/etc/prometheus/prometheus.yml \
    prom/prometheus

Lets inspect the data running a couple of errands in the Whip.

So its like 3:PM or so EST, and I needed to go to the store, along the way, I stopped at Taco Bell, then drove home to Gun Lake. Total trip was about 20 miles or so, with 2 stops.

Gun Lake -> Taco Bell -> Grocery Store -> Gun Lake

Distance to Empty

I hope you appreciate the simplicity here, but prometheus told the story of my errands run… I started out with a full charge with an extended setting, drove 10 miles, did two stops very close to each other, then drove it back 10 miles. You can even see where I did some drive way shuffling before I plugged it in to charge to Standard.

Gear Status

This one wont win any visualization awards, but if you recall the metric we did was 0 = P, and anything else is 1 = in motion (whether backwards or forwards). You can clearly see the 3 errands and the park shuffling in the gear status as well.

Everything you see here available on Github

https://github.com/sween/rivian-prometheus-exporter

References

Github: https://github.com/prometheus/client_python

Docker

Kubernetes