Notes
This release of the Python agent adds a deprecation warning for customers that disable SSL, and adds bugfixes for supported async frameworks that use asyncio.
The agent can be installed using easy_install/pip/distribute via the Python Package Index or can be downloaded directly from the New Relic download site.
Mandatory SSL
Disabling SSL connections to New Relic has been deprecated
SSL connections are enabled by default. In a future release, the option to disable SSL will be removed.
Bug Fixes
Using asyncio.gather or asyncio.ensure_future now tracks transaction context
Prior to this release, using asyncio.gather or asyncio.ensure_future may result in certain traces (such as external calls) not being reported in the transaction. Traces scheduled with asyncio.gather or asyncio.ensure_future from within the context of a transaction should now be properly attributed to the transaction.
Notes
This release of the Python agent contains improvements to the time tracing API as well as bug fixes.
The agent can be installed using easy_install/pip/distribute via the Python Package Index or can be downloaded directly from the New Relic download site.
Time Trace API Improvements
Time trace APIs (such as function_trace) can now be used with coroutines
The following decorator APIs can now be used with native coroutines and generators:
- function_trace
- database_trace
- datastore_trace
- external_trace
- message_trace
- memcache_trace
Example:
@function_trace(name='my_coroutine')async def my_coroutine():await asyncio.sleep(0.1)
Bug Fixes
gRPC instrumentation used on Python 2.x can cause a memory leak
When using gRPC on Python 2, gRPC futures would not be garbage collected resulting in a memory leak. gRPC futures will now be garbage collected.
Instrumentation for Dropbox v8.0 and newer caused error log messages
Dropbox client version 8.0 or higher raised instrumentation errors. These errors did not prevent metrics on Dropbox from being sent. These errors have been removed.
Values from negated ranges were sometimes added to ignore_status_codes
Negated status codes not found in the current ignore_status_codes were added if they were part of a range of values. This issue has been addressed.
Notes
This release of the Python agent contains security related updates as well as bug fixes.
The agent can be installed using easy_install/pip/distribute via the Python Package Index or can be downloaded directly from the New Relic download site.
Security Updates
See the associated security bulletins for custom API parameters in high-security mode and for SQL params for SQLite.
Bug Fixes
Using the aiohttp client may result in an application crash
Under certain circumstances, using the aiohttp client may have resulted in an application crash. This issue has been addressed.
Database queries made with psycopg2 may not have been recorded
When using the "with" statement to create a cursor, time spent on database calls may not have been properly recorded. This issue has been addressed.
Usage of the pika library resulted in a memory leak
When using the pika library with New Relic, Channel objects would not be cleared from memory as expected. This would result in abnormally high memory utilization in some cases. The memory leak has now been fixed.
DeprecationWarning generated for async/await keywords
Using the agent may have generated a DeprecationWarning around async/await being reserved keywords. The DeprecationWarning should no longer be generated.
Notes
This release of the Python agent adds improvements to django instrumentation, improvements to django rest framework instrumentation, and various bug fixes.
The agent can be installed using easy_install/pip/distribute via the Python Package Index or can be downloaded directly from the New Relic download site.
Improved Features
Enabled reporting of handled exceptions in Django REST Framework
Exceptions handled by django rest framework are now reported if the resulting response code is not ignored (see ignore_status_codes for details).
Add support for the error_collector.ignore_status_codes setting in Django
Ignoring exceptions in Django was previously limited to the error_collector.ignore_errors configuration option. Ignoring exceptions by response status code is now supported for Django through the use of the error_collector.ignore_status_codes configuration option.
Bug Fixes
Servicing aiohttp websocket requests results in an application crash
Servicing a websocket request in an aiohttp application may have resulted in an application crash when using the New Relic python agent. The application will now operate as expected when handling a websocket request.
Ignore incomplete aiohttp transactions
In aiohttp, connections can be terminated prior to the HTTP response being generated and sent. In those cases, the request handler may be cancelled. These transactions are no longer reported.
Fix HTTP status reporting for Tornado transactions
HTTP status was not properly added to Tornado transaction events and transaction traces. HTTP status is now automatically added to Tornado transaction events in Insights and transaction traces in APM.
Fix reporting of concurrent external requests in Tornado
External requests that execute in parallel in a tornado application may not have been recorded. This issue has been addressed.
Notes
This release of the Python agent adds instrumentation for aiohttp framework, adds support for displaying Heroku dyno names, and various bug fixes.
The agent can be installed using easy_install/pip/distribute via the Python Package Index or can be downloaded directly from the New Relic download site.
New Features
Add instrumentation for aiohttp framework
Data is now automatically collected for applications using the aiohttp framework. Data for aiohttp applications will appear in both APM and Insights. Additionally, cross application tracing is supported for incoming requests for aiohttp applications. In addition to service maps, aiohttp applications will now show the calling application in transaction traces.
Add support for displaying Heroku dyno names
Heroku-friendly logic is now applied to how dyno names are displayed. This includes being able to collapse dyno hostnames based on prefix.
Bug Fixes
Fix crash for gunicorn gaiohttp driver
Using gunicorn's gaiohttp worker with New Relic Browser monitoring enabled may have resulted in an application crash. This crash has been fixed and the gaiohttp worker is now fully supported with the New Relic Python Agent.
Fix crash for pika versions 0.9.x and earlier
Using the agent with pika versions 0.9.x and earlier could have resulted in an application crash. This issue has now been fixed.
Notes
This release of the Python agent adds instrumentation for aiohttp client.
The agent can be installed using easy_install/pip/distribute via the Python Package Index or can be downloaded directly from the New Relic download site.
New Features
Add instrumentation for aiohttp client
Outbound HTTP requests through the aiohttp library are now recorded. aiohttp Cross Application Tracing is now supported for outbound requests. In addition to Service Maps, applications accessed through the aiohttp client will now appear in transaction traces.
Bug Fixes
Fix crash when using psycopg2 v2.7 composable queries
The psycopg2 library introduced a module to generate SQL dynamically (psycopg2.sql) in version 2.7. Passing a Composable type object (psycopg2.sql.Composable) to execute or executemany resulted in an application crash. The agent now correctly handles psycopg2 Composable objects.
Notes
This release of the Python agent adds support for monitoring of outbound gRPC requests and adds an API for cross application tracing of custom non-HTTP transports. For details on APIs for cross application tracing see the cross application tracing document.
The agent can be installed using easy_install/pip/distribute via the Python Package Index or can be downloaded directly from the New Relic download site.
New Features
Add API for cross application tracing of non-HTTP external services
A new API is now exposed for implementing cross application tracing in custom instrumentation of non-HTTP transport libraries. For usage of this API see the cross application tracing document.
Add instrumentation for gRPC client calls
Outbound gRPC requests will now show up in APM under the External Services tab and in transaction traces.
Bug Fixes
Fixes erroneous recording of TastyPie
NotFound
exceptionsWhen a TastyPie API view raised a NotFound exception resulting in a 404 response, the agent may have erroneously recorded the exception. This has now been fixed.
Notes
This release of the Python agent includes bug fixes and improvements to the builtin pika instrumentation.
The agent can be installed using easy_install/pip/distribute via the Python Package Index or can be downloaded directly from the New Relic download site.
Improved Feature
Naming of Pika Message Transactions
This release improves the naming of message transactions . Message consumer transactions are now split based on the name of the consume callback.
Improved metadata collection for cloud providers
The agent now collects additional metadata when running in AWS, GCP, Azure, and PCF. This information is used to provide an enhanced experience when the agent is deployed on those platforms.
Bug Fixes
CherryPy WSGI Server automatic instrumentation
This release fixes a bug where
cherrypy.wsgiserver.CherryPyWSGIServer
was not being instrumented. This bugfix impacts CherryPy versions v6.1.0 to v8.9.1.
Notes
This release of the Python agent includes bug fixes for the pika RabbitMQ library instrumentation.
The agent can be installed using easy_install/pip/distribute via the Python Package Index or can be downloaded directly from the New Relic download site.
Bug Fixes
Using pika to make RPC-style calls from Tornado could result in an application crash
When making a remote procedure call from a tornado web application using the pika RabbitMQ library's Tornado connection, it is possible for the tornado web application to crash. Pika Tornado connections no longer result in an application crash.
AttributeError generated from an unregistered application starting a MessageTransaction
Applications which consume message transactions and do not call
register_application
can raise an AttributeError resulting in an application crash. This issue has now been fixed.
Notes
This release of the Python agent adds built-in instrumentation for the pika RabbitMQ library, a message broker API which can be used to add custom instrumentation for additional message brokers, and various bug fixes.
The agent can be installed using easy_install/pip/distribute via the Python Package Index or can be downloaded directly from the New Relic download site.
New Features
Pika Instrumentation
The RabbitMQ pika library is now automatically instrumented. The agent will report data for messages sent and received by an application. Data on messages is available in both APM and Insights. Applications connected through a RabbitMQ exchange will now also be visible on Service Maps as part of Cross Application Tracing.
Message Broker API
A Message Broker API is now available for use in custom instrumentation. The Message Broker API will send data to APM and Insights (through message.routingKey and message.queueName attributes) for message broker related operations.
Tornado 4.5 Support
The Python Agent's out-of-the-box Tornado 4 instrumentation now supports Tornado version 4.5.
Tornado AsyncIO Loop
In addition to the Tornado PollIOLoop, the AsyncIO loop is now supported as part of our Tornado instrumentation. When Tornado is configured to use the AsyncIO loop, data will now be sent to New Relic. Note that AsyncIO is not yet generally supported.
Bug Fixes
SQL comments beginning with a "--" resulted in an improperly parsed query
When executing SQL queries with comments that used the "--" comment syntax, the database operation and table was not properly parsed resulting in queries being absent from the Databases tab. Queries with "--" comments should now show up in the Database tab.
AttributeError generated when using Tornado 4 HTTPRequest objects in an external web call
When using the Tornado 4 HTTPRequest object as an argument to HTTPClient.fetch, it was possible to generate an AttributeError. The web call now completes as expected.
Slow network connections may result in lost data
When, due to a slow network, a 408 error is returned from New Relic data was not retransmitted. Data is now retransmitted when a 408 is received.