Here are some troubleshooting tips for custom instrumentation with New Relic's Java agent.
You can use New Relic's UI to define instrumentation rules with the Custom Instrumentation Editor, or edit your XML file via your Java app's Settings.
Java apps only: To edit your XML file directly from the New Relic UI: From the New Relic menu bar, select APM > (Applications) > (selected app) > Settings > Live Instrumentation. From here you can:
- Download a sample XML file.
- Select an edit existing XML file.
- Search the instrumentation history.
Nested transactions in thread profiler
Custom instrumentation is useful for nested transactions. For example, in the thread profiler, methods marked [circle] Instrumentation not allowed cannot be instrumented because this can result in unacceptable overhead. However, children of those methods usually can be instrumented.
To identify methods you can instrument, expand the thread profile tree until you find a suitable method:
- From the New Relic menu bar, select APM > Applications > (selected app) > Events > Thread profiler.
- Expand the thread profile tree until you find a suitable method.
- Follow standard procedures to define and deploy custom instrumentation.
Classes and methods
When troubleshooting custom instrumentation for your Java agent, compare the
pointcut information in your newrelic.yml config file with confirmation messages in your log file. Verify that the classes and methods match.
Here is an example:
# This is a pointcut example in your newrelic.yml config file: <pointcut transactionStartPoint="true"> <className>com.example.class.name</className> <method> <name>exampleMethod</name> </method> </pointcut>
If this is being instrumented properly, you may see a message similar to this in the log file:
# This is a confirmation example in your log file: Oct 1, 2015 10:58:52 -0700 [9805 1] com.newrelic FINER: Instrumenting com/example/class/name
If the log file class and method do not match your custom instrumentation values, review and adjust as needed. (You may not always see a confirmation in your log file about whether the custom instrumentation load succeeded or failed.)
Separate transaction in XML
Here is an example of a pointcut with several
method values. The
nameTransaction has been added to the XML to break out the method as a separate transaction rather than as a segment in the New Relic APM Transactions breakdown table.
# This is a pointcut example to identify a specific transaction with XML custom instrumentation: <pointcut> <nameTransaction/> <className>com.examplename.client.actionflow.impl.exampleActionFlow</className> <method> <name>requestNAME</name> <parameters> <type>boolean</type> </parameters> </method> </pointcut>
If your app makes use of async processes with custom instrumentation, it may not be possible to connect the async worker activity to the parent transaction. Async processes handle the web transaction in parallel. Many workers handle bits of the request and then reassemble the work to create the complete request.
Most agents do not tie these child transactions together as a single response. This is why you may not be able to identify the transaction in New Relic APM's user interface.
For more help
Additional documentation resources include: