Noteworthy changes to the agent are documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
- TBD
-
Support for Java 17 #433
-
Distributed Tracing is on by Default and deprecates Cross Application Tracing #486
- Increases the default maximum number of samples stored for Span Events from 1000 to 2000.
- The maximum number of samples stored for Span Events can be configured via the max_samples_stored configuration in the newrelic.yml.
span_events: max_samples_stored: 2000
Notice: This feature will cause an increase in the consumption of data. The amount depends on the application. This feature can be disabled by adding the following to the agent yaml config nested under the common stanza:
distributed_tracing: enabled: false
-
Auto-instrumentation Support for GraphQL-Java 17.0+ #487
-
This version tested agent support for the ARM64/Graviton2 platform
The existing MongoDB sync client instrumentation was incorrectly applying when MongoDB reactive or async client was being used, which could lead to segment timeouts and long transaction response times. #476
Cross Application Tracing is now deprecated, and disabled by default. To continue using it, enable it with cross_application_tracer.enabled = true
and distributed_tracing.enabled = false
.
- New Relic recommends that you upgrade the agent regularly to ensure that you're getting the latest features and performance benefits. Additionally, older releases will no longer be supported when they reach end-of-life.
-
Support for Java 16
-
Auto-instrumentation support for java.net.http.HttpClient
-
Migrate the Agent’s caching library from Guava to Caffeine
- Caffeine provides an in-memory cache using a Google Guava inspired API. The improvements draw on the author’s experience designing Guava's cache and
ConcurrentLinkedHashMap
. - We expect this change to provide improvement in cases where we saw thread contention and deadlocks attributable to the Guava library.
- Caffeine provides an in-memory cache using a Google Guava inspired API. The improvements draw on the author’s experience designing Guava's cache and
- Removed support for the anorm-2.0 instrumentation module
- The artifacts that this module instrumented are no longer available.
- New Relic recommends that you upgrade the agent regularly to ensure that you're getting the latest features and performance benefits. Additionally, older releases will no longer be supported when they reach end-of-life.
-
Scala Library Instrumentation #362 and #363
- STTP versions 2 & 3 Akka-HTTP, HTTP4s and STTP core backends
- Cats-effect v2
- ZIO v1
- HTTP4s client & server v0.21
- Play 2.3-2.8
- Akka-HTTP v10.1 & v10.2
- For more information, see Scala instrumentation.
-
Scala API support (see PRs above)
- Scala APIs provided for explicit instrumentation of several of above libraries in case auto-instrumentation is not desired
- Cats-effect v2
- ZIO v1
-
AWS v2 DynamoDB Instrumentation #343
- Synchronous and asynchronous AWS v2 APIs are auto-instrumented similarly to v1 APIs
- For more information, see Add support for AWS SDK 2 DynamoDB sync/async clients
-
GraphQL 16 Instrumentation #396
- Create meaningful transaction names
- Create meaningful spans
- Reporting GraphQL errors
- For more information, see GraphQL for Java
-
JFR feature causing excessive overhead when enabled JFR #203
- Refactored code to use less memory.
The existing MongoDB instrumentation was partially applying when MongoDB Reactive Streams is being used.
- Disable weaving package when MongoDB 4.x+ reactive driver detected #341
- For more information, see Spring Reactive DB Drivers - MongoDB Support
- New Relic recommends that you upgrade the agent regularly to ensure that you're getting the latest features and performance benefits. Additionally, older releases will no longer be supported when they reach end-of-life.
Due to overhead caused in some applications Real-time profiling for Java using JFR metrics is now disabled by default.
It can be enabled following the instructions here.
Some customers saw increased overhead when Real-time profiling is enabled. See instructions to disable Real-time profiling in the notice below.
-
Java instrumentation by XML new properties #288
- traceLambda - to trace lambdas inside a method
- traceByReturnType - to trace all methods in a class that return a given type
These are compatible with Java and Scala. For more information, see Java instrumentation by XML.
-
Scala APIs #254
New artifacts allow Scala code to be instrumented using a fluent Scala API instead of the Java annotations. There are specific artifacts for versions 2.10, 2.11, 2.12, 2.13 of Scala. Scala 3.0 users can use the 2.13 artifact.
For more information, see Scala instrumentation.
-
Real-time profiling for Java using JFR metrics #333
Real-time profiling for Java using JFR metrics is now enabled by default.
Notice: this feature will cause an increase in the consumption of data. The amount depends on the application. It can be disabled by adding the following to the agent yaml config nested under the
common
stanza:jfr: enabled: false
For more information, see JFR core README.
- Fixes an issue where the agent would break OkHttp versions 3.X and lower. #324
-
Real-time profiling for Java using JFR metrics
Real-time profiling for Java using JFR metrics is now fully integrated into the Java agent. See the JFR core README for additional details.This feature requires a supported version of Java (Java 8 (specifically version
8u262
+) or Java 11+) and is currently disabled by default. To enable it set the following in your yaml (indented 2 spaces under the common stanza).jfr: enabled: true
Notice: If you were previously using the jfr-daemon jar as an agent extension or standalone process you should remove that option to avoid potential conflicts with the JFR service that is now built into the agent.
-
Not compatible with Java 7
In order to continue to innovate and efficiently provide new capabilities to our customers who run on the JVM, this and future agent versions are not compatible with Java 7. If you are running Java 7, you may continue to use Java agent 6.5.0 or lower. For details, see this topic on the Explorers Hub. -
Adds support for akka http with Scala 2.13 #271
-
Class annotation to trace lambda methods #274
-
Class annotation to trace methods by return type #275
-
Fixes an issue that could cause multiple versions of
akka-http-core
instrumentation to apply at the same time. #208 -
The agent will now log dropped events at
FINE
instead ofWARN
to decrease verbosity. #296 -
Fixes Javadoc comments that incorrectly stated that, when calling the
noticeError
API multiple times, the first error would be reported when in fact it is the last error that is reported. #313
New Relic recommends that you upgrade the agent regularly to ensure that you're getting the latest features and performance benefits. Additionally, older releases will no longer be supported when they reach end-of-life.
- The agent no longer bundles SSL certificates with it and the
use_private_ssl
option that configured the agent to use the previously bundled certificates has been removed. By default, the agent will use the SSL truststore provided by the JVM unless it is explicitly configured to use a different truststore with the ca_bundle_path option. See Configuring your SSL certificates for more details. (#245)
- Fixes an issue that could cause incorrect transaction naming when using JAX-RS sub-resources. (#234)
- Reactor Netty instrumentation improvements and fixes. (#237, #239, #243)
- Java 7 compatibility deprecation
In order to continue to innovate and efficiently provide new capabilities to our customers who run on the JVM, Java 7 support has been deprecated and this will be the last version of the agent compatible with it.
If you are running Java 7, you may continue to use Java agent 6.5.0 or lower.
For more information, see the Explorers Hub post.
- New Relic recommends that you upgrade the agent regularly to ensure that you're getting the latest features and performance benefits. Additionally, older releases will no longer be supported when they reach end-of-life.
- Spring Webflux/Netty Reactor instrumentation improvements for enhanced tracing across asynchronous thread hops (#174, #190, #195).
- Infinite tracing will now utilize a backoff sequence on retries. (#180)
- New distributed tracing APIs have been added to better support general use cases for
propagating distributed tracing headers. In particular the new APIs provide enhanced support for W3C Trace Context but
are flexible enough to support other header protocols. Previous distributed tracing APIs have been deprecated and are subject to removal in a
future agent release. See documentation here.
(#188)
Transaction.insertDistributedTraceHeaders(Headers)
is used to create and insert distributed tracing headers (both newrelic and W3C Trace Context) into aHeaders
data structure.Transaction.acceptDistributedTraceHeaders(TransportType, Headers)
is used to accept the distributed tracing headers sent from the calling service and link these services together in a distributed trace.
- Updated the Java agent’s snakeyaml dependency to 1.27. (#182)
- In some environments the jar collector service could lead to high CPU utilization at application startup. The agent now provides a configurable rate limiter, with a reasonable default, for processing jars detected in the application’s environment. See documentation here. (#183)
- New Relic recommends that you upgrade the agent regularly to ensure that you're getting the latest features and performance benefits. Additionally, older releases will no longer be supported when they reach end-of-life.
- Adds support for Spring Webflux 5.3.+.
- Adds configuration for ignoring netty reactor errors. The relevant configuration property is
reactor-netty.errors.enabled
. Error reporting is enabled by default. - Adds support for Scala 2.13, including Akka Http and Play. Thank you junder31 for these contributions.
- The netty-4.0.8 instrumentation would sometimes not start a Transaction on
channelRead
, potentially affecting instrumentation dependent on it including: Spring, Akka and Play. - Updates the Java agent’s Apache HttpClient dependency to 5.13.
- Spring Webclient could report the wrong URL when multiple HTTP calls to several URLs occurred in parallel. Thank you veklov for contributing this fix.
- New Relic recommends that you upgrade the agent regularly to ensure that you're getting the latest features and performance benefits. Additionally, older releases will no longer be supported when they reach end-of-life.
- Support for Real Time Profiling of your JVMs. The agent can now run in parallel with the JFR Daemon to provide Realtime Profiling of your JVMs using Java Flight Recorder! Read more about it in our blog and documentation.
- The agent now supports parsing cgroup ids that do not contain
docker-
. Previously, the agent only supported docker cgroup ids that containeddocker-
in them, such as1:cpu:/system.slice/docker-12345abcde.scope
. The agent now supports parsing cgroup ids such as1:cpu:/system.slice/crio-12345abcde.scope
. - Adds support for Akka Http Core 10.2.0. Thank you jobicarter for reporting this issue.
- Adds support for gRPC 1.30.0. Thank you jef for submitting this request and trying it out.
- Adds support for
map
andflatmap
async external calls in spring webflux framework. Previously the agent didn’t support client calls that occurred within themap
orflatmap
methods within the spring webflux framework. External calls such asserviceB
andserviceC
in the example below would not get reported to New Relic:
return serviceA.getData()
.map(service -> Response.builder().service(service).build())
.flatMap(serviceB::getData)
.flatMap(serviceC::getData)
.doOnNext(this::saveResponse);
- Adds support for Play 2.16.3. Many thanks to junder31 for this contribution.
- Prevents the agent from logging a timeout exception when using New Relic Edge with Infinite Tracing but the agent hasn’t sent spans in a while.
- Prevents the agent from logging a socket exception when trying to retrieve cloud provider information in a non-cloud environment.
- Adds New Relic EU certifications if
ca_bundle_path
is specified. This fix also came with the reintroduction of theuse_private_ssl
config, which can be set to add our agent certs to the truststore.
- New Relic recommends that you upgrade the agent regularly to ensure that you're getting the latest features and performance benefits. Additionally, older releases will no longer be supported when they reach end-of-life.
- Support for Java 15
- Ability to add attributes to segments
Newrelic.addCustomParameter()
API now supports boolean values- Fix a config issue where the agent would try to read environment variables using system-property syntax. It will now log the correct syntax and ignore the incorrect config
- The Java agent now includes the newrelic.com SSL certificate.
In previous agent versions, applications using a custom Truststore would have to provide their certificate or use the
use_private_ssl
configuration which was removed in 6.0.0. - Fixed an issue where applications could fail to start due to the agent attempting to access the JMX MBean server before it was initialized.
- Fix for asynchronous jar file collection. Collection of modules will no longer block the initial first harvest.
- Support for okhttp 4.4 instrumentation.
- Fix for
reactor-netty
verifier versions - Improve reconnect behavior by not pinning to preconnected collector host instance.
- Fix race condition around connection pool exhaustion by separating pool timeout from connection timeout.
- Register an MBean to expose the agent linking metadata.
Final closed-source release.
See the external release notes.