Quarkus issue reproduction: quarkus-opentelemetry clears MDC context of ManagedExecutor thread

quarkusio/quarkus#43134

Project to reproduce an issue observed in a Quarkus based project. In short: the behaviour observed is that for logging statements executed on a ManagedExecutor thread, the MDC (traceId) context is not available.

Project setup

This project was created using the quarkus CLI:

quarkus:create
quarkus ext add quarkus-opentelemetry

• See GreetingResource for the code.
• See application.properties for the configuration.
  • The mdc.traceId is added to logging statements using the quarkus.log.console.format configuration.

Running reproduction

./mvnw quarkus:dev

curl localhost:8080/hello

Observed behaviour

2024-09-09 10:12:49,785 INFO  traceId=e19013fa3e2b7e0ce8365af3149e09c0 [org.acm.GreetingResource] (executor-thread-2) hello() [traceId=e│
19013fa3e2b7e0ce8365af3149e09c0]                                                                                                        │
2024-09-09 10:12:49,787 INFO  traceId=e19013fa3e2b7e0ce8365af3149e09c0 [org.acm.GreetingResource] (executor-thread-1) hello() from Manag│
edExecutor before sleep [traceId=e19013fa3e2b7e0ce8365af3149e09c0]                                                                      │
2024-09-09 10:12:50,788 INFO  traceId= [org.acm.GreetingResource] (executor-thread-1) hello() from ManagedExecutor after sleep [traceId=│
e19013fa3e2b7e0ce8365af3149e09c0]

Observations:

• In the ManagedExecutor thread, the mdc.traceId is sometimes not included in the logging statement.
• On the contrary, tracing context (Span.current().getSpanContext().getTraceId()) is not cleared, thus always correctly available in the ManagedExecutor thread
• My investigation/understanding of the issue:
  • Vert.x duplicated context is used in the ManagedExecutor thread
  • VertxMDC is used as MDCProvider, stored in the Vert.x context
  • quarkus-opentelemetry clears the MDC data in this duplicated context (via OpenTelemetryUtil#clearMDC(...))