Python

Reference
2-min read

You can send data from your Python application to Dynatrace. Python is a versatile, high-level programming language known for its readability and simplicity, often used for web development, data analysis, artificial intelligence, and scientific computing due to its extensive libraries and community support.

Capabilities

Injects into Python processes and supported containers automatically. Injection rules enable the exclusion or inclusion of specific processes and containers.
Tracks CPU usage, memory consumption, and responsiveness of Python processes in real time.
Monitors Python workloads running inside containers, capturing container-specific performance metrics and metadata about the container technology and environment.
Provides runtime metrics, including garbage collection (GC) activity, thread metrics, and loaded library versions.
Provides remote configuration updates that allow monitoring settings to be adjusted remotely via OneAgent. Changes are applied at runtime (without requiring a process restart when possible).
Generates diagnostic data for troubleshooting, including support archives accessible via the OneAgent Diagnostics UI.
Comprehensive end-to-end transaction tracing for requests to web services, remote services, message queues, databases, and others. For more information, see Services.
Third-party vulnerability detection. For more information, see Runtime Vulnerability Analytics.
OpenTelemetry support for capturing traces. For more information, see Instrument your Python application with OpenTelemetry.
OneAgent SDK for custom tracing.

If you already use the OneAgent SDK for Python or OneAgent SDK Python auto-instrumentation, we recommend migrating to enable auto-instrumentation using OneAgent for Python.

See our supported technologies matrix for details on supported technologies used in conjunction with Python.

Supported Python versions

Python version

Vendor released

Vendor End of life

First supported Dynatrace OneAgent version

Last supported Dynatrace OneAgent version

Dynatrace support until

Dynatrace support level

3.13

2024-10-07

2029-10-01

1.309

Supported

3.12

2023-10-02

2028-10-01

1.309

Supported

3.11

2022-10-24

2027-10-01

1.309

Supported

3.10

2021-10-04

2026-10-01

1.309

Supported

3.9

2020-10-05

2025-10-01

1.309

Supported

3.8

2019-10-14

2024-10-07

1.309

Supported

Enable Python monitoring

To activate Python monitoring

Go to Settings > Monitoring > Monitored technologies.
Find Python in the list of supported technologies, and select Edit.
Turn on Monitor Python.
Create a process monitoring rule to enable deep monitoring of the selected processes.

Process logs with technology bundle parsers

Through OpenPipeline, you can use and configure technology bundles. A technology bundle is a library of parsers (processing rules), that process logs from various technologies such as Java, .NET, Microsoft IIS, etc.

Parsers help you to improve filtering, troubleshooting, metrics, alerts, and dashboards by efficiently extracting log levels and relevant attributes. You can also use technology bundles to structure logs from technologies that are not supported by Dynatrace out of the box.

For more information, see Process logs with technology bundle parsers.

OneAgent Python code module and OneAgent SDK for Python compatibility

The OneAgent Python code module supports applications instrumented by the OneAgent SDK for Python. Both can work together in the same application.
The Dynatrace OneAgent SDK for Python is a wrapper of the Dynatrace OneAgent SDK for C/C++.

When the Python code module is installed, it replaces Dynatrace OneAgent SDK for C/C++ internally and collects data produced by OneAgent SDK for Python. This means that traces produced by the SDK will be linked with those produced by the Python code module itself.

Method hotspots

The Python code module can capture Python stack traces in the background for the method hotspots feature.

Prerequisites

Requires Python version 3.12+.
The OneAgent Python code module must be switched to non-forkable mode. To do this, set the environment variable DT_PYTHON_FORKABLE to 0. After setting the environment variable, restart the application.

Limitations

The application will be multi-threaded even during forking. Ensure that forks do not execute any Python code without first calling the executable.
Samples are taken asynchronously and program execution is not stopped. Therefore, stacks may be incomplete or even contain incorrect call-by associations.

Limitations

Only the standard CPython interpreter is supported (the implementation from Python.org).
No-GIL (nogil) builds are not supported. The binary names of the nogil version have the t suffix, for example, Python3.13t.
PEP 703 - Making the Global Interpreter Lock Optional in Python is not supported.
Monitoring of processes with Gevent package installed in versions lower than 20.9.0 is not supported.
For applications creating forked child processes: The code module uses background threads. To support applications using process forking, the code module usually ensures that its background threads are shut down before a fork and restarted afterwards. The following limitations apply:
- On Python 3.12 and newer, the Python interpreter will log a warning like DeprecationWarning: This process (pid=12345) is multi-threaded, use of fork() may lead to deadlocks in the child. This is because after the fork and before Python checks if the process is multi-threaded, the code module already has to restart its threads. If the warning only appears with the Python code module and not if it is removed, it is a false positive and can be ignored. Otherwise, it may indicate a problem in application code and should be taken seriously.
- When forking without invoking Python fork handlers, the code module does not shut down its threads. This makes the parent process a multi-threaded process, with all the usual consequences this has for forking. For typical applications this is not relevant, but custom native extension modules calling lower-level forking APIs directly would be affected.