Stefan Wehner created LOG4J2-620:
------------------------------------
Summary: Deadlock on reconfiguration with Appenders that use log4j
Key: LOG4J2-620
URL: https://issues.apache.org/jira/browse/LOG4J2-620
Project: Log4j 2
Issue Type: Bug
Components: Configurators
Affects Versions: 2.0-rc1
Reporter: Stefan Wehner
We're using the JDBC appender to log to database and provide connections from a
C3P0 pool.
We're finding deadlocks when reconfiguring the log4j2 configuration under heavy
load.
It seems like C3P0 writes to log4j when getting/returning connections from/to
the pool and the application deadlocks.
I think what happens is: When the appender gets a connection from the pool, the
call to log4j from C3P0 notices the configuration change, tries to reconfigure,
and waits for all appenders (including itself) to finish...
To simplify things I've written a small sample application (Maven project) with
a custom appender that uses log4j in its append method.
As you can see the application deadlocks soon (usually on the first
reconfiguration)
--
This message was sent by Atlassian JIRA
(v6.2#6252)
---------------------------------------------------------------------
To unsubscribe, e-mail: log4j-dev-unsubscr...@logging.apache.org
For additional commands, e-mail: log4j-dev-h...@logging.apache.org
