apache
investigate ClusterActorLoggingSpec
possibly related to #771
- passes on laptop both for 1.0.x and main branches
- failed in nightly build but passed when rerun
- seems to pass when nightly build is run using the 1.0.x branch - https://github.com/apache/incubator-pekko/actions/runs/6902135347/job/18778244515
still flaky in nightly builds: https://github.com/apache/pekko/actions/runs/9993956065/job/27622606159
Because sometimes the actor don't put value into MDC:
LoggingEvent(
INFO,
org.apache.pekko.cluster.typed.ClusterActorLoggingSpec,
ClusterActorLoggingSpec-pekko.actor.default-dispatcher-11,
Starting,
1721630292397,
None,
None,
+ Map(pekkoAddress -> pekko.tcp://[email protected]:42905, pekkoSource -> pekko://ClusterActorLoggingSpec/user/$a, sourceActorSystem -> ClusterActorLoggingSpec)
)
LoggingEvent(
INFO,
org.apache.pekko.cluster.typed.ClusterActorLoggingSpec,
ClusterActorLoggingSpec-pekko.actor.default-dispatcher-3,
Starting,
1721631006935,
None,
None,
- Map()
)
Because sometimes the actor don't put value into MDC
Doesn't that mean the test caught a bug?
Doesn't that mean the test caught a bug?
It is flaky depending on how fast SLF4J initialization.
If it loads slower, it will use SubstituteLogger(which uses BasicMDCAdapter) as a fallback.
if we use MDC earlier than SLF4J initialization, the MDCAdapter may no chance to update it.
won't reopen it because ci is passing: https://github.com/apache/pekko/actions/runs/10076913173/job/27858277926