Prometheus remote storage adapter for Dynatrace

This is an experiment to use Dynatrace as remote storage for Prometheus. It implements both, the remote_read as well as the remote_write functionality for Prometheus.

Possible Usecases:

• Query and combine data transparently through Prometheus
• Forward Prometheus-collected data to Dynatrace
• Design Grafana dashboards with Dynatrace data using PromQL
• Use the remote_write and remote_read implementation in Python as inspiration for other projects

Architecture Overview

Getting Started

Using Docker

A minimal example can be launched using docker compose:

• Create a file named .env with your credentials:

DT_TENANT='tov08230.sprint.dynatracelabs.com'
DT_API_TOKEN='dt0c01.NS2DS75F...'

• Adjust the docker-compose.yml file to your needs (optional)
• run docker from the project root folder:

docker-compose up

Prometheus UI will be available on http://localhost:9090. There you can query for Dynatrace metrics using an adapted syntax where the __name__ label represents the metric you're looking for (see limitations).

{__name__="builtin:synthetic.http.duration.geo"}

Manual

You can run it on your own:

### If you use venv
# python3 -m pip install --user virtualenv
# python3 -m venv env
# source env/bin/activate

python3 -m pip install -r requirements.txt

You'll need to configure your tenant url and API token (API v2 with the metrics read/write capabilities). This can be done as ENV variables or inside an .env file.

export DT_TENANT=<your_tenant_url>
export DT_API_TOKEN=<your_apiv2_token>

Then you're good to go:

python3 -m flask run

Prometheus Configuration

Point your Prometheus instance to this application. See documentation for further configuration options.

Eg. if your Prometheus is running in a docker container use:

remote_read:
 - url: http://host.docker.internal:5000/read

remote_write:
- url: http://host.docker.internal:5000/write
  queue_config:
    max_samples_per_send: 1000
  metadata_config:
    send: true

Limitations

There are some limitation:

Dyntrace vs Prometheus metric naming

Metric and label/dimension naming rules between Dynatrace and Prometheus are varying slightly. Therefore following conversion will be applied:

., , and - will be replaced by _ on remote read (from Dynatrace)

: will be replace by _ on remote write (to Dynatrace)

If you try to query Prometheus for an invalid metric name like builtin:synthetic.http.duration.geo (as . is not an allowed character) you'll see a message like:

Error executing query: invalid parameter "query": 1:18: parse error: unexpected character: '.'

As a workaround use one of the alternative syntaxes:

• Embed the metricname as the name label: {__name__="builtin:synthetic.http.duration.geo"}
• replace all unsupported chars with underscores (_) - The adapter will look up the actual metric name: builtin:synthetic_http_duration_geo

Prometheus does not list remote metric names

As of now you can not get autocomplete of Dynatrace metrics in the UI. This is a known limitation for remote storage and is being addressed in prometheus/prometheus#7076.