prometheus-lightsail-sd

Service discovery for the AWS Lightsail platform, compatible with Prometheus.

❗ Deprecation Notice ❗

This project has been integrated into upstream as Prometheus native service discovery:

Thanks to the Prometheus Dev Team.

How it works

This service gets the list of servers from the Lightsail API and generates a file which is compatible with the Prometheus file_sd mechanism.

Pre-requisites

1) IAM Policy

When using AWS credentials or IAM Roles, the following policy needs to be attached to the object:

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Effect": "Allow",
            "Action": [
                "lightsail:GetInstances"
            ],
            "Resource": [
                "*"
            ]
        }
    ]
}

2) AWS Credentials

After attaching the policy above to the target object, authenticate using one of the three options:

2a) AWS Named Profile

To use an AWS named profile, your profile config should be defined under ~/.aws/config and ~/.aws/credentials.

The profile name can be specified with either:

Command line argument --profile=myProfileName
Setting the environment variables AWS_PROFILE=myProfileName & AWS_REGION=us-east-1

More info: AWS CLI - Named Profiles

2b) AWS IAM Key via Environment Variables

Set AWS IAM Key values via environment variables:

Set the environment variables:
- AWS_ACCESS_KEY_ID
- AWS_SECRET_ACCESS_KEY=EXAMPLEKEY
- AWS_REGION=us-east-1

2c) IAM Role attached to instance

This tool supports AWS IAM Roles attached to instances, ie. running in EC2/ECS/etc.

Note: Lightsail instances do not support IAM Roles, so you must use steps 2a or 2b.

Installing it

Download the binary from the Releases page.

Running it

usage: prometheus-lightsail-sd [<flags>]

Tool to generate file_sd target files for AWS Lightsail.

Flags:
  -h, --help               Show context-sensitive help (also try --help-long and --help-man).
      --output.file="lightsail_sd.json"  
                           Output file for file_sd compatible file.
      --target.refresh=60  The refresh interval (in seconds).
      --profile=""         AWS Profile
      --web.listen-address=":9888"  
                           The listen address.
      --web.telemetry-path="/metrics"  
                           Path under which to expose metrics.
      --version            Show application version.

Example output:

$ ./prometheus-lightsail-sd --profile=myProfileName
level=debug ts=2019-11-20T08:56:30.089673118Z caller=main.go:175 msg="loading profile: myProfileName"
level=debug ts=2019-11-20T08:56:30.089986153Z caller=manager.go:224 msg="Starting provider" provider=lightsailSD subs=[lightsailSD]
level=debug ts=2019-11-20T08:56:31.355158911Z caller=main.go:117 msg="get servers" count=2
level=debug ts=2019-11-20T08:56:31.355313346Z caller=main.go:123 msg="server added" source=lightsail/sccbc.net
level=debug ts=2019-11-20T08:56:31.355479273Z caller=main.go:123 msg="server added" source=lightsail/n888.net

Prometheus /service-discovery example output

Systemd service file example

See examples/systemd/prometheus-lightsail-sd.service

Integration with Prometheus

Here is a Prometheus scrape_config snippet that configures Prometheus to scrape node_exporter (TCP/9100) on discovered instances:

- job_name: node

  # This config assumes that prometheus and prometheus-lightsail-sd are started from the same directory:

  file_sd_configs:
  - files: [ "./lightsail_sd.json" ]

  # The relabeling does the following:
  # - overwrite the scrape address with the node_exporter's port.
  # - add the value of the instance's custom tag named "service".
  # - add the availability zone label.
  # - overwrite the instance label with the server's name.
  
  relabel_configs:
  - source_labels: [__meta_lightsail_private_ip]
    replacement: "${1}:9100"
    target_label: __address__
  - source_labels: [__meta_lightsail_tag_service]
    target_label: service
  - source_labels: [__meta_lightsail_availability_zone]
    target_label: availability_zone
  - source_labels: [__meta_lightsail_name]
    target_label: instance

The following meta labels are available on targets during relabeling:

__meta_lightsail_availability_zone="us-west-2a": availability zone
__meta_lightsail_blueprint_id="ubuntu_18_04": blueprint id (lightsail pre-baked image)
__meta_lightsail_bundle_id="nano_2_0": instance size
__meta_lightsail_instance_id="i-00f8cb1818387aa7z": instance id
__meta_lightsail_name="n888.net": instance name
__meta_lightsail_private_ip="172.88.3.6" instance private ip
__meta_lightsail_public_ip="50.33.32.72" instance public ip
__meta_lightsail_state="running" instance state
__meta_lightsail_support_code="782236961567/i-00f8cb1818387aa7z" instance support code
__meta_lightsail_tag_service="frontend" instance tag, each tag gets its own label

note: lightsail.Instance does not return an InstanceId (unlike EC2), but does provide a "support code"
string with the format of "${lightsail_account_id}/${instance_id}". This string is split and the
${instance_id} value is assigned to the `instance_id` meta label.

Building

promu is used for building (https://github.com/prometheus/promu):

Build current arch:

promu build

Build all archs under .promu.yml:

promu crossbuild

Contributing

Pull requests, issues and suggestions are appreciated.

Credits

Prometheus Authors
Scaleway Service Discovery
- Core code based on https://github.com/scaleway/prometheus-scw-sd
aws-sdk-go
- https://aws.amazon.com/sdk-for-go

License

Apache License 2.0, see LICENSE.

n888/prometheus-lightsail-sd