Community
Participate
Working Groups
need support to handle asynchronous events in model for Webservices instrumentation in application tracing using ARM. The trace model needs to support these events. At the user interface level, negative time values to appear causing incorrect UML sequence diagrams. For ARM monitored applications in a async webservice (or RMI) this needs to be fixed for a true distributed application.
Fixed version and target milestone to be consistent with TPTP development process. For defects, version specified is the version where the defect was found. The target milestone is the version that has or will have the fix for it.
143814 is a general large undertaking for TPTP, I defer this to future until we can consider a design in a future major release of TPTP. However, raising as P1 as it is an important scenario to cover.
Scrubbing some old defects... Harm - I recall you mentioning that there is a way to handle async properly in the current Trace model. I must be losing my mind because I can't remember how this is done. Can you please add some notes here to refresh my memory please?
Sorry for the delay on this, but at the modeling level this should simply be an invocation event that spans threads or even processes and machines. In an abstract sense, that is all this is, and then each of the threads continues as usual. The truth is that we never really capture every single thing going on, and thread managers, dispatch queues and so on are rarely in the trace. There for you are basically indicating an abstraction that shows resulting control flow. The next layer of issues is how viewers and analyzers deal with this in the model. Not just form a visualization standpoint, but also issues such as time accumulation, and the fact the invocation never returns; just as an example. Web Services, workflows etc. in the end all run on top of plumbing that supports these distributed concepts, and in all cases the lowest level of detail is normally hidden and the abstractions need to be "rendered".
From my tests in TPTP 441 async representation seems to be working. Negative time values as initially discussed should not have been appearing and has been fixed in the BTM piece.
verified and closing.
