He-Pin opened a new issue, #3687:
URL: https://github.com/apache/pekko/issues/3687
## Summary
`startDefaultLoggers` resolves and creates each configured logger inside a
single for-comprehension. If entry *N* cannot be resolved, entries *1..N-1*
have already been created as system actors, but the `loggers` field is only
assigned after the whole comprehension succeeds — so the teardown path never
sees them.
## Evidence
`actor/src/main/scala/org/apache/pekko/event/Logging.scala:137-159`:
```scala
val myloggers =
for {
loggerName <- defaultLoggers
if loggerName != StandardOutLogger.getClass.getName
} yield {
system.dynamicAccess
.getClassFor[Actor](loggerName)
.map { actorClass =>
addLogger(system, actorClass, level, logName) // creates the
actor
}
.recover {
case e =>
throw new ConfigurationException(
"Logger specified in config can't be loaded [" +
loggerName +
"] due to [" + e.toString + "]", e)
}
.get // throws here
}
guard.lock()
try {
loggers = myloggers // :159 —
never reached
```
`addLogger` (`:216-226`) creates the actor and blocks on initialization:
```scala
val actor =
system.systemActorOf(Props(clazz).withDispatcher(system.settings.LoggersDispatcher),
name)
implicit def timeout: Timeout = system.settings.LoggerStartTimeout
...
try Await.result(actor ? InitializeLogger(this), timeout.duration)
```
`stopDefaultLoggers` (`:192-209`) iterates the `loggers` field:
```scala
for {
logger <- loggers
if logger != StandardOutLogger
} {
// this is very necessary, else you get infinite loop with DeadLetter
unsubscribe(logger)
logger match {
case ref: InternalActorRef => ref.stop()
case _ =>
}
}
```
## Impact
With `pekko.loggers = [<valid>, <invalid>]`, the valid logger is created as
a system actor and initialized, then the `ConfigurationException` propagates
before `loggers = myloggers` executes. System creation fails, and
`stopDefaultLoggers` iterates a `loggers` value that does not contain the
already-started actor — so it is neither unsubscribed nor stopped. A
partially-initialized logging setup outlives the failed `ActorSystem` creation.
## Suggested direction
Resolve every class name first and fail fast before creating any actor, then
create the loggers. Alternatively, accumulate successfully-created loggers so
the failure path can unsubscribe and stop them.
## Test
`actor-tests/src/test/scala/org/apache/pekko/event/LoggerSpec.scala` — a
case configuring one valid plus one unresolvable logger, asserting that system
creation fails with `ConfigurationException` and that the valid logger actor
has been stopped and unsubscribed.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
