AWS Prometheus Exporter

This Python module allows you to run AWS API calls through Boto3, and expose the results of those calls as Prometheus metrics. Metrics must be described in YAML. For example:

ec2_instance_ids:
  description: EC2 instance ids
  service: ec2
  paginator: describe_instances
  paginator_args:
    Filters:
      - Name: instance-state-name
        Values: [ "running" ]
  label_names:
    - instance_id
  search: |
    Reservations[].Instances[].{instance_id: InstanceId, value: `1`}

The above describes a Prometheus metric derived from calling the following in Boto3:

result = boto3.client("ec2") \
            .get_paginator("describe_instances") \
            .paginate({"Filters": [{"Name": "instance-state-name", "Values": ["running"]}]}) \
            .search("Reservations[].Instances[].{instance_id: InstanceId, value: `1`}")

You aren't restricted to calling paginators. By specifying method and method_args in place of paginator and paginator_args, you can call any service method. Pagination will be handled for you in the following fashion:

  service = boto3.client("ec2")
  next_token = ''
  result = []
  kwargs = {"Filters": [{"Name": "instance-state-name", "Values": ["running"]}]}
  while next_token is not None:
      response = boto3.client("ec2").describe_instances(**kwargs)
      next_token = response.get('NextToken', None)
      result += jmespath.search("Reservations[].Instances[].{instance_id: InstanceId, value: `1`}", response)
      kwargs["NextToken"] = next_token  

The dict values returned by the paginator or service method are then converted by this module into GaugeMetricFamily samples. Each dict must have the same keys as label_names, plus an additional value key; the corresponding values correspond to the labels and value of the created Gauge, respectively.

Note that pagninator_args may be a string. In that case, it will be eval-ed with access to datetime.datetime and datetime.timedelta. For example:

recent_emr_cluster_ids:
  description: Recent EMR cluster ids
  service: emr
  paginator: list_clusters
  paginator_args: |
    {
        "CreatedAfter": datetime.now() - timedelta(weeks=4)
    }
  label_names:
    - id
  search: |
    Clusters[].{id: Id, value: `1`}

Example Usage

The module can be run directly, as follows:

python -m aws_prometheus_exporter --metrics-file ./metrics.yaml --port 9000 --period-seconds 300

Running using Docker:

docker build -t aws_prometheus_exporter:latest  .
docker run -p9000:9000 -v $(pwd)/example.yaml:/mnt/metrics.yaml aws_prometheus_exporter

Alternatively, you can import the module into an existing application. See __main__.py for an example.

Links

• Boto3 docs: https://boto3.amazonaws.com/v1/documentation/api/latest/index.html
• Boto3 paginator docs: https://boto3.amazonaws.com/v1/documentation/api/latest/guide/paginators.html
• Prometheus: https://prometheus.io/
• Prometheus Python Client: https://github.com/prometheus/client_python