Debug ID in MDC

[mabn]

An example of keeping Span's debug_id inside MDC - for version 0.31.0+

Puts active Spans's debug_id when scope is activated and cleans it
afterwards.

see also: #206

[coveralls]

Coverage remained the same at 78.598% when pulling 8230a90 on futuresimple:debug_id_in_mdc into cdc9601 on opentracing:v0.31.0.

[mabn]

Real implementation will also require casting so other libs that wrap Span will make it fail :|

[coveralls]

Coverage remained the same at 78.598% when pulling 8230a90 on futuresimple:debug_id_in_mdc into cdc9601 on opentracing:v0.31.0.

[carlosalberto]

Maybe call it something other than debugId so you don't have to use this.debugId for the instance field? (same name but different types, etc)

[carlosalberto]

Out of curiosity: is this something you would foresee being implemented as a wrapper class? I ask because if the user does scopeManager.active().close(), the logic you used for MDC will not be called, i.e.

            } finally {
                 MDC.put(mdcKey, previousDebugId);
            }

[mabn]

Oh, you're right. I should pass my MdcScope when calling wrapped.activate(..)

If this wouldn't be a wrapper then it wouldn't be possible to use it with another scope manager (e.g. AutoFinishScopeManager). Until there is an API to register observers/listeners/extensions then wrapping is the only way :|

[sudr]

Wondering when this feature will become available.

[whiskeysierra]

@mabn Isn't this what https://github.com/opentracing-contrib/java-api-extensions is for?

[yurishkuro]

I haven't tried api-extensions, but subclassed the scope manager to add trace IDs to logs.

https://github.com/PacktPublishing/Mastering-Distributed-Tracing/blob/master/Chapter11/exercise1/lib/src/main/java/lib/MDCScopeManager.java

[whiskeysierra]

As long as 0.32.0 is not released there is no way to get this done without being implementation-specific, right?

https://github.com/PacktPublishing/Mastering-Distributed-Tracing/blob/master/Chapter11/exercise1/lib/src/main/java/lib/MDCScopeManager.java#L29

[yurishkuro]

@whiskeysierra yeah. Even with 0.32, the traceSampled field cannot be logged generically.

[mabn]

@whiskeysierra yes, java-api-extensions would be suitable, but IIRC at the time they didn't have some of the callbacks I needed

I'm closing this PR as I haven't worked on this for quite a while.