That is correct, but if you only log to files that may not be so bad. If you use any asynchronous stuff or things that buffer events then that would be a problem.
Ralph > On Jul 22, 2015, at 2:41 AM, Mikael Ståldal <mikael.stal...@magine.com> wrote: > > Yes, Spark seems to have its own shutdown hook and it logs stuff from > within there. > > If I put shutdownHook="disable" in my log4j2.xml, the problem goes away. > But then I guess that Log4j is never shut down properly? > > On Wed, Jul 22, 2015 at 8:39 AM, Ralph Goers <ralph.go...@dslextreme.com> > wrote: > >> The currentContext value indicates whether the LoggerContext stored in the >> ThreadLocal should be used. This is usually the wrong choice which is why >> the value is usually set to false. Instead, the LoggerContext is located >> via the ClassLoaderContextSelector which finds the LoggerContext for the >> ClassLoader of the class that owns the Logger. >> >> It looks to me like something is trying to log after logging has already >> shut down. This is causing it to retrieve the LoggerContext, which has >> already been stopped, and then it tries to restart it. Because the >> LoggerContext is normally stopped by the execution of the shutdown hook it >> is a very bad sign that the shutdown hook is trying to be reestablished. >> My guess is that Spark has its own shutdown hook and is trying to log from >> it after our shutdown hook has already run. >> >> When Matt wrote this code I believe he allowed for this situation but I’d >> really have to dig into the code more to figure out. However, I did find >> http://stackoverflow.com/questions/17400136/how-to-log-within-shutdown-hooks-with-log4j2 >> < >> http://stackoverflow.com/questions/17400136/how-to-log-within-shutdown-hooks-with-log4j2> >> which says to implement your own ShutdownCallbackRegistry. >> >> It might be a good idea to add the variation of the Shutdown registry >> mentioned in the article so that it can be configured. >> >> Ralph >> >>> On Jul 21, 2015, at 8:57 PM, Remko Popma <remko.po...@gmail.com> wrote: >>> >>> Looks like the context returned by >>> AbstractLoggerAdapter.java:102 >>> is the result of LogManager.getContext(someClass, false); >>> >>> The boolean "false" indicates that a different LoggerContext may be >> created (and started). I suspect this is what causes the problem. >>> >>> Ralph understands this part better than me. Ralph, should the slf4j >> Log4jLoggerFactory pass "false" here? The default for LogManager seems to >> be "true"... >>> >>> Remko >>> >>> Sent from my iPhone >>> >>>> On 2015/07/21, at 22:58, Mikael Ståldal <mikael.stal...@magine.com> >> wrote: >>>> >>>> AbstractLoggerAdapter.java:102 >>> >>> --------------------------------------------------------------------- >>> To unsubscribe, e-mail: log4j-user-unsubscr...@logging.apache.org >>> For additional commands, e-mail: log4j-user-h...@logging.apache.org >>> >>> >> >> > > > -- > [image: MagineTV] > > *Mikael Ståldal* > Senior backend developer > > *Magine TV* > mikael.stal...@magine.com > Regeringsgatan 25 | 111 53 Stockholm, Sweden | www.magine.com > > Privileged and/or Confidential Information may be contained in this > message. If you are not the addressee indicated in this message > (or responsible for delivery of the message to such a person), you may not > copy or deliver this message to anyone. In such case, > you should destroy this message and kindly notify the sender by reply > email. --------------------------------------------------------------------- To unsubscribe, e-mail: log4j-user-unsubscr...@logging.apache.org For additional commands, e-mail: log4j-user-h...@logging.apache.org
