Datadog Python APM Client

ddtrace is Datadog’s Python APM client. It is used to profile code and trace requests as they flow across web servers, databases and microservices. This enables developers to have greater visibility into bottlenecks and troublesome requests in their application.

Getting Started

For a basic product overview: check out the setup documentation.

For details about developing and contributing: refer to the development guide.

For descriptions of the terminology of Datadog APM: take a look at the official documentation.

Supported Libraries

We officially support Python 2.7, 3.5 and above.

The versions listed are the versions that we have tested, but ddtrace can still be compatible with other versions of these libraries. If a version of a library you use is unsupported, feel free to contribute or request it by contacting support.

Integration	Supported Version	Automatically Instrumented 1
aiobotocore	>= 0.2.3	No
aiohttp	>= 2.0	Yes 2
aiopg	>= 0.12.0	Yes
algoliasearch	>= 1.20.0	Yes
asgi	>= 2.0	No
asyncio		Yes 5
boto2	>= 2.29.0	Yes
botocore	>= 1.4.51	Yes
Bottle	>= 0.11	No
Celery	>= 3.1	Yes
Cassandra	>= 3.5	Yes
CherryPy	>= 11.2.0	No
Consul	>= 0.7	Yes 3
dogpile.cache	>= 0.6	Yes
Django	>= 1.8	Yes
djangorestframework	>= 3.4	No
Elasticsearch	>= 1.6	Yes
Falcon	>= 1.0	No
Fastapi	>= 0.51	Yes
Flask	>= 0.10	Yes
Flask Cache	>= 0.12	No
gevent	>= 1.1	No
Grpc	>= 1.12.0	Yes
Jinja2	>= 2.7	Yes
Mako	>= 0.1.0	Yes
MariaDB	>= 1.0.0	Yes
kombu	>= 4.0	No
Molten	>= 0.7.0	Yes
Mongoengine	>= 0.11	Yes
mysql-connector	>= 2.1	No
mysqlclient	>= 1.3	No
psycopg	>= 2.7	Yes
pylibmc	>= 1.4	Yes
Pylons	>= 0.9.6	No
pymemcache	>= 1.3	Yes
Pymongo	>= 3.0	Yes
PynamoDB	>= 4.0	Yes
PyODBC	>= 4.0	Yes
Pyramid	>= 1.7	No
pytest	>= 4.0.0	Yes
redis	>= 2.6	Yes
redis-py-cluster	>= 1.3.5	Yes
Requests	>= 2.08	Yes
Sanic	>= 19.6.0	Yes 4
SQLAlchemy	>= 1.0	No
Starlette	>= 0.13.0	Yes
Tornado	>= 4.0	No
urllib3	>= 1.22	No
Vertica	>= 0.6	Yes

1

Libraries that are automatically instrumented when the ddtrace-run command is used or the patch_all() method is called. Always use patch() and patch_all() as soon as possible in your Python entrypoint.

2

only third-party modules such as aiohttp_jinja2

3

only the synchronous client

4

only in Python 3.7 and above.

5

Activating the legacy context provider is required in Python < 3.7. See asyncio for more details.

Indices and tables

• Index

• Module Index

• Search Page