Skip to content
This repository has been archived by the owner on Jul 25, 2024. It is now read-only.

dynatrace-oss/opentelemetry-metric-python

Dynatrace

Dynatrace supports native OpenTelemetry protocol (OTLP) ingest for traces, metrics and logs. All signals can be sent directly to Dynatrace via OTLP protobuf over HTTP using the built-in OTLP/HTTP Exporter available in the OpenTelemetry Python SDK. More information on configuring your Python applications to use the OTLP exporter can be found in the Dynatrace documentation.

Dynatrace OpenTelemetry Metrics Exporter for Python

Static Badge

Warning Dynatrace supports native OpenTelemetry protocol (OTLP) ingest for traces, metrics and logs. Therefore, the proprietary Dynatrace OpenTelemetry metrics exporter is deprecated in favor of exporting via OTLP/HTTP.

The exporter is still available but after the end of 2023, no support, updates, or compatibility with newer OTel versions will be provided.

Please refer to the migration guide for instructions on how to migrate to the OTLP HTTP exporter, as well as reasoning and benefits for this transition.

For an example on how to configure the OTLP exporter in a Python application, check out the Python integration walk-through page in the Dynatrace documentation.

This exporter allows exporting metrics created using the OpenTelemetry SDK for Python directly to Dynatrace.

It was built against OpenTelemetry SDK version 1.12.0 and should work with any 1.12+ version.

More information on exporting OpenTelemetry metrics to Dynatrace can be found in the Dynatrace documentation.

Getting started

Installation

To install the latest version from PyPI run:

pip install opentelemetry-exporter-dynatrace-metrics

Usage

from opentelemetry import metrics
from opentelemetry.sdk.metrics import MeterProvider
from dynatrace.opentelemetry.metrics.export import (
    configure_dynatrace_metrics_export
)


# setup metrics export pipeline
metrics.set_meter_provider(MeterProvider(
    # configure Exporter/MetricReader combination with a 5000ms export
    # interval, endpoint url and API token.
    metric_readers=[
        configure_dynatrace_metrics_export(
            export_interval_millis=5000,
            endpoint_url=endpoint_url,
            api_token=api_token)
    ]))

# get a meter
meter = metrics.get_meter(__name__)

# create a counter instrument and provide the first data point
counter = meter.create_counter(
    name="my_counter",
    description="Description of MyCounter",
    unit="1"
)

counter.add(25, {"dimension-1": "value-1"})

Example

To run the example, clone this repository and change to the opentelemetry-metric-python folder, then run:

pip install .           # install the Dynatrace exporter
pip install psutil      # this package is used by the example to read CPU/Memory usage
export LOGLEVEL=DEBUG   # (optional) Set the log level to debug to see more output (default is INFO)
python example/basic_example.py

A more complete setup routine can be found here, including installing inside a virtual environment and getting required packages. If you just want to see it in action, it should be sufficient to run example/install_and_run.sh from the root folder. This script will install Python, set up a virtual environment, pull in all the required packages and run the example.

The example also offers a simple CLI. Run python example/basic_example.py -h to get more information.

Configuration

The exporter allows for configuring the following settings by passing them to the constructor:

Dynatrace API Endpoint

The endpoint to which the metrics are sent is specified using the endpoint_url parameter.

Given an environment ID myenv123 on Dynatrace SaaS, the metrics ingest endpoint would be https://myenv123.live.dynatrace.com/api/v2/metrics/ingest.

If a OneAgent is installed on the host, it can provide a local endpoint for providing metrics directly without the need for an API token. This feature is currently in an Early Adopter phase and has to be enabled as described in the OneAgent metric API documentation. Using the local API endpoint, the host ID and host name context are automatically added to each metric as dimensions. The default metric API endpoint exposed by the OneAgent is http://localhost:14499/metrics/ingest. If no endpoint is set and a OneAgent is running on the host, metrics will be exported to it automatically using the OneAgent with no endpoint or API token configuration required.

Dynatrace API Token

The Dynatrace API token to be used by the exporter is specified using the api_token parameter and could, for example, be read from an environment variable.

Creating an API token for your Dynatrace environment is described in the Dynatrace API documentation. The permission required for sending metrics is Ingest metrics (metrics.ingest) and it is recommended to limit scope to only this permission.

Metric Key Prefix

The prefix parameter specifies an optional prefix, which is prepended to each metric key, separated by a dot (<prefix>.<name>).

Default Dimensions

The default_dimensions parameter can be used to optionally specify a list of key/value pairs, which will be added as additional dimensions to all data points. Dimension keys are unique, and labels on instruments will overwrite the default dimensions if key collisions appear.

Export Dynatrace Metadata

If running on a host with a running OneAgent, setting the export_dynatrace_metadata option to True will export metadata collected by the OneAgent to the Dynatrace endpoint. If no Dynatrace API endpoint is set, the default exporter endpoint will be the OneAgent endpoint, and this option will be set automatically. Therefore, if no endpoint is specified, a OneAgent is assumed to be running and used as the export endpoint for all metric lines, including metadata. More information on the underlying Dynatrace metadata feature that is used by the exporter can be found in the Dynatrace documentation.

Dimensions precedence

When specifying default dimensions, attributes and Dynatrace metadata enrichment, the precedence of dimensions with the same key is as follows: Default dimensions are overwritten by attributes passed to instruments, which in turn are overwritten by the Dynatrace metadata dimensions (even though the likeliness of a collision here is very low, since the Dynatrace metadata only contains Dynatrace reserved dimensions starting with dt.*).

Development

Requirements

Just tox. Make sure to pip install the requirements-dev.txt to get the relevant packages.

Running tests and lint

  • Test all supported python versions: tox
  • Test all supported python versions in parallel: tox -p
  • A particular python version: tox -e 38
  • Current python version: tox -e py
  • Lint: tox -e lint

Limitations

Typed attributes support

The OpenTelemetry Metrics API for Python supports the concept of Attributes. These attributes consist of key-value pairs, where the keys are strings and the values are either primitive types or arrays of uniform primitive types.

At the moment, this exporter only supports attributes with string key and value type. This means that if attributes of any other type are used, they will be ignored and only the string-valued attributes are going to be sent to Dynatrace.

Histogram

OpenTelemetry Histograms are exported to Dynatrace as statistical summaries consisting of a minimum and maximum value, the total sum of all values, and the count of the values summarized. If the min and max values are not directly available on the metric data point, estimations based on the boundaries of the first and last buckets containing values are used.