Since 3.10 the interlok-profiler has an additional aspect available to it (you should be able to use it with previous versions of Interlok since this is a pure aspect with few dependencies) which adds the messageId and the parentMessageId to your logging context (the parentMessageId might be useful if you are using a splitter). Since logging contexts are thread-based, depending on where in your processing chain you are and what service combinations your workflow contains, you can lose the logging context or gain an unwanted logging context. With this in mind, you can view the profiling logging context as a way of guaranteeing that you have the correct logging context via the use of an AspectJ aspect rather than being frustrated because your workflow design isn’t suitable for add-logging-context-service or logging-context-workflow-interceptor
Getting started
The easiest way is to use dependency management to and include interlok-profiler as a dependency; once you have the appropriate jar file then you can start Interlok with the appropriate runtime configuration.
JAVA_ARGS="-javaagent:./lib/aspectjweaver.jar -Dorg.aspectj.weaver.loadtime.configuration=META-INF/profiler-logging-context.xml"
java $JAVA_ARGS -jar lib/interlok-boot.jar
Note that there are 3 profiler configurations available in the interlok-profiler jar file. You can of course write your own if your requirements don’t match what’s available
- META-INF/profiler-aop.xml - this provides performance metrics data that you can use with prometheus
- META-INF/profiler-logging-context.xml - Just includes the aspect for adding the logging context
- META-INF/profiler-interlok-all.xml - All of the above.
What it does
The code itself is very simple actually, and it’s probably easier to just have it here in its entirety:
public static final String MESSAGE_ID_CONTEXT = "messageId";
public static final String PARENT_ID_CONTEXT = "parentMessageId";
@Before("(call(* com.adaptris.core.Service+.doService(com.adaptris.core.AdaptrisMessage)) "
+ "|| call(* com.adaptris.core.AdaptrisMessageListener+.onAdaptrisMessage(com.adaptris.core.AdaptrisMessage)) "
+ "|| call(* com.adaptris.core.AdaptrisMessageListener+.onAdaptrisMessage(com.adaptris.core.AdaptrisMessage, java.util.function.Consumer))) ")
public synchronized void beforeService(JoinPoint jp) {
AdaptrisMessage msg = (AdaptrisMessage) jp.getArgs()[0];
String msgId = msg.getUniqueId();
String parentId = msg.getMetadataValue(CoreConstants.PARENT_UNIQUE_ID_KEY);
MDC.put(MESSAGE_ID_CONTEXT, msgId);
if (StringUtils.isNotEmpty(parentId)) {
MDC.put(PARENT_ID_CONTEXT, parentId);
}
}
@Before("call(* com.adaptris.core.PollerImp+.processMessages())")
public synchronized void beforeProcessMessages(JoinPoint jp) {
MDC.remove(MESSAGE_ID_CONTEXT);
MDC.remove(PARENT_ID_CONTEXT);
}
- Before every invocation of
doService(), onAdaptrisMessage()
within concrete implementations of Service or AdaptrisMessageListener add the current message’s uniqueId and if available, the parentMessageId against the keysmessageId
andparentMessageId
respectively. - Before every invocation of
processMessages()
within a concrete subclass of PollerImp, remove the contextual information.
Log4j2 configuration
You can now use %X to grab contextual information out of the diagnostic context and use that as part of your log4j2 pattern.
<Console name="Console" target="SYSTEM_OUT">
<PatternLayout>
<Pattern>%-5p [%t] [%c{1.}] [%X{parentMessageId}].[%X{messageId}] %m%n</Pattern>
</PatternLayout>
</Console>
which leads to logging like this (since this particular example isn’t inside a splitter, we don’t get anything for the parentMessageId
)
TRACE [JMX-Request-0:fervent-tesla.START] [c.a.c.u.LifecycleHelper] [].[] Executing Start(): EmbeddedConnection(jetty)
TRACE [JMX-Request-0] [c.a.c.u.LifecycleHelper] [].[] Executing Start(): NoRetries(jolly-ptolemy)
TRACE [JMX-Request-0] [c.a.c.u.LifecycleHelper] [].[] Executing Start(): SharedComponentList
INFO [JMX-Request-0] [c.a.c.Adapter] [].[] Adapter(MyInterlokInstance) Started
DEBUG [gigantic-davinci@fervent-tesla] [c.a.c.StandardWorkflow] [].[868e716f-e960-4c68-be5f-94fd099b15d6] start processing msg [DefaultAdaptrisMessageImp[uniqueId=868e716f-e960-4c68-be5f-94fd099b15d6,metadata=[[jettyURI]=[/api/logging], [httpmethod]=[GET], [jettyURL]=[http://localhost:8080/api/logging], [_interlokMessageConsumedFrom]=[/api/logging]]]]
DEBUG [gigantic-davinci@fervent-tesla] [c.a.c.ServiceList] [].[868e716f-e960-4c68-be5f-94fd099b15d6] Executing doService on [EmbeddedScriptingService(select-failure)]
DEBUG [gigantic-davinci@fervent-tesla] [c.a.c.ServiceList] [].[868e716f-e960-4c68-be5f-94fd099b15d6] Executing doService on [IfElse(shall-i-fail)]
TRACE [gigantic-davinci@fervent-tesla] [c.a.c.s.c.IfElse] [].[868e716f-e960-4c68-be5f-94fd099b15d6] Running logical 'IF', with condition class ConditionMetadata
TRACE [gigantic-davinci@fervent-tesla] [c.a.c.s.c.c.ConditionMetadata] [].[868e716f-e960-4c68-be5f-94fd099b15d6] Testing metadata condition with key: failure
TRACE [gigantic-davinci@fervent-tesla] [c.a.c.s.c.IfElse] [].[868e716f-e960-4c68-be5f-94fd099b15d6] Logical 'IF' evaluated to true, running service.
DEBUG [gigantic-davinci@fervent-tesla] [c.a.c.ServiceList] [].[868e716f-e960-4c68-be5f-94fd099b15d6] Executing doService on [AlwaysFailService(failure-triggered)]
The project used to create the logging extract can be found on github