Notes
This release of the Python agent adds support for the following:
- Custom tasks in Celery
- Graphene-Django monitoring
- Middleware filtering in Django
- AWS Bedrock Converse API
- W3CTraceParent header sampled flag
This release also provides the following fixes:
- Safeguards for the deepest unique path in GraphQL
- Safeguards for web URL request parsing
- Fallback for ElasticSearch and GraphQL when
trace.settings==None
- Remove
options
method from async ElasticSearch client list
Install the agent using easy_install/pip/distribute
via the Python Package Index or download it directly from the New Relic download site.
New features
-
Adds support for custom tasks in Celery
-
Adds support for Graphene-Django
- Add monitoring support for synchronous schemas created with Graphene-Django. Previously, schemas created with Graphene-Django would only be partially monitored, resulting in missing GraphQL related attributes.
-
Adds support for middleware filtering in Django
-
Users are now able to filter which Django middleware to monitor. There are now three additional settings:
instrumentation.middleware.django.enabled
(Default istrue
)instrumentation.middleware.django.exclude
instrumentation.middleware.django.include
-
-
Adds support for AWS Bedrock Converse API
- Add monitoring support for non-streaming Converse API. This includes support for
converse
chat completion calls made usingboto3
andaioboto3
clients.
- Add monitoring support for non-streaming Converse API. This includes support for
-
Adds support for W3CTraceParent header sampled flag
- When an upstream service sends the W3CTraceParent header with the sampling flag set, New Relic will take into account that sampling decision when deciding whether the current transaction will be sampled.
-
Added support for sampling decisions based on whether the remote parent is sampled
-
Added support for two new configuration options:
distributed_tracing.sampler.remote_parent_sampled
distributed_tracing.sampler.remote_parent_not_sampled
-
default
is the default behavior and behaves as it did before this change (it passes the sampling decision to the adaptive sampling algorithm to determine whether the transaction will be sampled). -
always_on
means that the trace that has a remote parent sampled or remote parent not sampled will always be sampled. -
always_off
means that the trace that has a remote parent sampled or remote parent not sampled will never be sampled.
-
Bug fixes
-
Add safeguard for deepest unique path in GraphQL
- Some users were experiencing issues with the deepest unique path in GraphQL not being generated due to a field name returning
None
. This issue has been fixed.
- Some users were experiencing issues with the deepest unique path in GraphQL not being generated due to a field name returning
-
Add safeguard for web request parsing
- Some users experienced a parsing error when
request_uri
was malformed or contained invalid characters. This can happen at this point if therequest_uri
in the request header is invalid and does not match the (valid) URL originally given. While this safeguard has now been put in place in New Relic,urllib
has resolved this issue for versions of Python released after Jan 31, 2025.
- Some users experienced a parsing error when
-
Fix crash in ElasticSearch and GraphQL when
trace.settings==None
- In cases where
trace.settings
isNone
, the instrumentation for ElasticSearch crashed the application. Aglobal_settings
fallback has been put in place. This also resolves the same issue found in GraphQL.
- In cases where
-
Remove
options
method from async ElasticSearch client list- The previous instrumentation of ElasticSearch caused a crash because
options
was in the list of async methods. This method is no longer wrapped in aDatastoreTrace
since it is not one that conducts any datastore operations/requests.
- The previous instrumentation of ElasticSearch caused a crash because
Support statement
We recommend updating to the latest agent version as soon as it's available. If you can't upgrade to the latest version, update your agents to a version no more than 90 days old. Read more about keeping agents up to date.
See the New Relic Python agent EOL policy for information about agent releases and support dates.