lucaslorentz / durabletask-extensions / 28682323254

77%

push

github

fix(efcore): support DTFx distributed tracing without breaking execution

Enabling DurableTask.Core distributed tracing
(CorrelationSettings.Current.EnableDistributedTracing = true) hung every
orchestration on the EFCore backend instead of merely fragmenting spans.

Root cause: the backend never populated TaskOrchestrationWorkItem.TraceContext.
When tracing is on, TaskOrchestrationDispatcher seeds
CorrelationTraceContext.Current from workItem.TraceContext (null) and then
dereferences it unconditionally
(ExecutionStartedEvent.Correlation = CorrelationTraceContext.Current.SerializableTraceContext),
throwing a NullReferenceException. The work item is aborted and retried
forever, so the orchestration never completes. This legacy App-Insights
correlation path runs even under the W3CTraceContext protocol.

Fix: in LockNextTaskOrchestrationWorkItemAsync, restore the work item's trace
context from the ExecutionStartedEvent's correlation payload
(TraceContextBase.Restore, which returns a valid empty context when there is
none), mirroring the reference backends. Guarded on EnableDistributedTracing so
the tracing-off default stays zero-overhead.

The W3C span carriers (ExecutionStartedEvent.ParentTraceContext /
TaskScheduledEvent.ParentTraceContext) already round-trip through the
Newtonsoft serializer, so no schema, column, or serialization changes are
needed — the entire bug was the null TraceContext.

Adds an acceptance test (InMemory + Postgres + SqlServer + MySql variants)
asserting execution completes and that client -> orchestration -> activity
spans all share the caller's root trace id.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

9 of 9 new or added lines in 1 file covered. (100.0%)

2299 of 3002 relevant lines covered (76.58%)

141.88 hits per line