On Apr 5, 2007, at 12:17 PM, mehmet bilgin wrote:

i am using log4j with jboss application server.the smtpappender doesnot fulfill my needs so i write another smtpappender.this appender has an attribute name String mailTo.so you can change the mailto address from the xml file.i have an asyncappender which contains a reference to the my smtpappender.since jboss also uses log4j and it has a log4j.xml file i write my loggers in that file.and as you know jboss has a thread that checks whether this file changed or not. think of a position that asyncappender has 5 event in its queue. at that moment if i make a change in the log4j.xml file jboss thread tries to update themailTo field of my appender. But since there is an ongoing event in that appender this thread is get blocked.if again at that moment i send another event to my appender it also gets blocked (so my sessionbean gets blocked).what happens is first the 5 events in the queue is processed and then the mailTo field updated and finallymy session bean finishes the logging event.
what i want is my session bean never gets blocked. how can i do that.


Since you are using blocking=false on AsyncAppender, I assume that you are using a very recent log4j (either 1.2.13 or .14 if I remember correctly). However, I don't think that is directly related to your issue.

I would try to find a different way to communicate the change of email address instead of modifying log4j.xml. I'm speculating since I've never dug into the sequence of events that occur when a configuration is changed. My speculation that a modification of log4j.xml would block almost all of log4j and would force all the old appenders to complete any pending tasks (like processing the rest of the logging events in the queue) before rebuilding the configuration. That would be consistent with the behavior you observed. I don't think (but could be wrong) that there is logic to detect only the changes from the previous configuration and to not touch unmodified appenders. If your new log4.xml file contains an AsyncAppender, then I believe the old one must be closed (which would block until the queue is finished) before it is replaced with an newly constructed one.


---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to