PAYARA-4123 Refactored server JAXRS Request Tracing #4219
Conversation
dmatej
commented
Sep 16, 2019
dmatej
commented
- removed filter + exception mapper solution, which failed TCK's completitioncallback tests
- introduced EventListener solution, depends on Jersey
- removed trailing whitespaces in the module
- added logging and several generics
- reduced nesting of blocks in favor of readibility
- EnvironmentConfigSource uses AccessController to avoid security exceptions
|
Jenkins test please |
There was a problem hiding this comment.
Haven't fully reviewed the code yet, but made some observations.
I'm not convinced this works quite right - or it at least works differently.
I get different trace outcomes from simply enabling request tracing with the following settings:
- Applications Only = Disabled
- Threshold Unit = Milliseconds
- Selected Notifiers = service-log
https://gist.github.com/Pandrex247/fde4ed9fa2c813e2a4359e8f2dca18b9
https://gist.github.com/Pandrex247/80d0b0dda3f8419877d778cc7baed7a4
I also get an IOException when running the default endpoint of the following app from a blog I did a while ago, with it claiming that the active scope is null due it already being shut, something that worked previously. The example is a bit weird admittedly, but should probably be looked at.
Also also, I'm getting these log messages printed out which look a bit like debug statements:
onEvent(event.type=MATCHING_START)
onEvent(event.type=REQUEST_FILTERED)
|
|
|
Jenkins test please |
There was a problem hiding this comment.
NPE when enabling request tracing from admin console:
- Enable request tracing, Threshold Unit Microseconds, Applications Only disabled, service-log notifier enabled.
- Values save successfully but NPE in logs:
java.lang.NullPointerException
at fish.payara.notification.requesttracing.RequestTrace.assignReferences(RequestTrace.java:307)
at fish.payara.notification.requesttracing.RequestTrace.endTrace(RequestTrace.java:184)
at fish.payara.nucleus.requesttracing.RequestTraceSpanStore.endTrace(RequestTraceSpanStore.java:88)
at fish.payara.nucleus.requesttracing.RequestTracingService.endTrace(RequestTracingService.java:428)
|
&& span.getTraceEndTime().compareTo(comparisonSpan.getTraceEndTime() Could it be currently processed request? |
|
I didn't get conflicts when I merged them locally... |
- removed filter + exception mapper solution, which failed TCK's completitioncallback tests - introduced EventListener solution, depends on Jersey - removed trailing whitespaces in the module - added logging and several generics - reduced nesting of blocks in favor of readibility - EnvironmentConfigSource uses AccessController to avoid security exceptions
- tracedAnnotation instance cannot be resolved if the event is for example ON_EXCEPTION invoked sooner than after REQUEST_MATCHED event. - how to simulate: invoke unmapped context causing HTTP 404.
|
It depends on merging strategy and GIT settings - it detected "file copy" but it was false positive. I will push rebased version after TCK tests ... |
|
Jenkins test please |
