Skip to content

Latest commit

 

History

History
144 lines (109 loc) · 6.04 KB

README.md

File metadata and controls

144 lines (109 loc) · 6.04 KB
Raw

statsbomb/prometheus-msk-discovery Test Go Report Card

prometheus-msk-discovery

Service discovery for AWS MSK, compatible with Prometheus.

How it works

This service gets a list of MSK clusters in an AWS account and exports each broker to a Prometheus-compatible static config to be used with the file_sd_config or http_sd_config mechanism.

Pre-requisites

1) IAM Policy

When using AWS credentials or IAM Roles, the following policy needs to be attached to the role/user being used:

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Effect": "Allow",
            "Action": [
                "kafka:ListClusters",
                "kafka:ListNodes"
            ],
            "Resource": [
                "*"
            ]
        }
    ]
}

2) AWS Credentials

  • No special provisions are made to obtain AWS credentials and instead that process is left to the AWS SDK. Credentials will be searched for automatically in the order specified in the documentation.

Running it

Usage of ./prometheus-msk-discovery:
  -filter string
    	a regex pattern to filter cluster names from the results
  -http-sd
    	expose http_sd interface rather than writing a file
  -job-prefix string
    	string with which to prefix each job label (default "msk")
  -listen-address string
    	Address to listen on for http service discovery (default ":8080")
  -output string
    	path of the file to write MSK discovery information to (default "msk_file_sd.yml")
  -region string
    	the aws region in which to scan for MSK clusters
  -scrape-interval duration
    	interval at which to scrape the AWS API for MSK cluster information when in file_sd mode (default 5m0s)
  -tag value
    	A key=value for filtering by tags. Flag can be specified multiple times, resulting OR expression.

Example output:

$ ./prometheus-msk-discovery -scrape-interval 10s -filter 'primary'
2021/04/04 21:02:55 Writing 1 discovered exporters to msk_file_sd.yml

An example output file can be found here

http_sd

$ ./prometheus-msk-discovery -http-sd -listen-address :8989 -filter 'primary'

$ curl localhost:8989
[{"targets":["b-1.primary-kafka.tffs8g.c2.kafka.eu-west-2.amazonaws.com:11001","b-1.primary-kafka.tffs8g.c2.kafka.eu-west-2.amazonaws.com:11002","b-2.primary-kafka.tffs8g.c2.kafka.eu-west-2.amazonaws.com:11001","b-2.primary-kafka.tffs8g.c2.kafka.eu-west-2.amazonaws.com:11002"],"labels":{"job":"msk-primary-kafka","cluster_name":"primary-kafka","cluster_arn":"arn:aws:kafka:eu-west-2:111111111111:cluster/primary-kafka/522d90ab-d400-4ea0-b8fd-bbf3576425d4-2"}}]

http_sd_configs:
  - url: http://localhost:8989
    refresh_interval: 30s

Region Precedence

When no region is specified with the -region flag the process first attempts to load the default SDK configuration checking for an AWS_REGION environment variable or reading any region specified in the standard configuration file. If no region is found it will attempt to retrieve it from the EC2 Instance Metadata Service.

Explicitly setting a region with the -region flag takes precedence over everything else.

Integration with kube-prometheus-stack

The Docker image for this project can be used to inject a container into the Prometheus Spec of a kube-prometheus-stack installation by using the following snippet in values.yaml:

prometheus:
  prometheusSpec:
    containers:
      - name: prometheus-msk-discovery
        image: statsbomb/prometheus-msk-discovery:latest
        args:
          - -output
          - /config-out/msk_file_sd.yml
        volumeMounts:
          - name: config-out
            mountPath: /config-out

You'll then need to add something similar to the following to your additionalScrapeConfig:

- job_name: prometheus-msk-discovery
  file_sd_configs:
    - files:
        - /etc/prometheus/config_out/msk_file_sd.yml
  honor_labels: true

Other Kubernetes Setups

If you're not using the kube-prometheus-stack Helm chart then the general idea for running prometheus-msk-discovery is that it needs to be run as a sidecar container alongside the container that is running Prometheus. You'll need to ensure that there is a shared volume mounted to both of the containers in order for Prometheus to be able to read the config file that prometheus-msk-discovery writes.

Building

Building can be done just by using go build

Contributing

Pull requests, issues and suggestions are appreciated.

Credits

License

Apache License 2.0, see LICENSE.