logging-log4j-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Francesco Chicchiriccò <ilgro...@apache.org>
Subject Re: Using Spring-managed DataSource (or other Spring beans) in log4j2 configuration
Date Mon, 16 Sep 2013 15:46:47 GMT
On 13/09/2013 19:29, Nick Williams wrote:
> On Sep 13, 2013, at 10:26 AM, Francesco Chicchiriccò wrote:
>> Hi all,
>> I am trying to configure a JDBCAppender in log4j 2.0-beta9-SNAPSHOT, with a pre-existing
DataSource managed via Spring; such DataSource is used by the rest of Syncope.
>> I have tried, without success. to provide a custom ConnectionSource or a ConnectionFactory:
the point is that, when log4j2 initializes, the Spring context is yet to be built, hence no
beans (including the DataSource referenced above) are available.
>> I have read from [1] that log4j2 features an "automatic configuration": is there
any way to disable it and invoke it manually later?
> You don't want to do this, because you could miss important logging events. What happens
if there is some failure in Spring that doesn't get logged because you delayed configuration?
You always want Log4j to start before anything else in your application; otherwise, you could
miss important events.

Ok, got the point.

> Could you tell me a little bit more exactly how your DataSource is "spring-managed?"
This isn't exactly clear to me. Are you actually using Spring Framework to create the DataSource
somehow? If so, what Spring Framework feature/classes are you using? If Spring Framework has
to start in order to create your DataSource, this is going to be a problem because logging
will start too late.
> There is definitely a way to get this working, and I can help you, I just need a little
better understanding of exactly how you're creating and distributing the DataSource.

We have a "dataSource" bean (check full definition [2]) empowering the 
defaultObject feature of Spring's JndiObjectFactoryBean [3]: e.g., any 
datasource access will first attempt to find the specified JNDI 
resource, otherwise it will go with the fallback (inefficient, 
not-for-production) DriverManagerDataSource.
Such datasource is used ubiquitously in Syncope core.

After your reply, I have developed a simple ConnectionFactory for use 
with JDBCAppender that basically does the same job as 
JndiObjectFactoryBean, reading the same configuration data and I am 
quite satisfied of the result.

Anyway, if it could be of any interest, I've found the following 
non-blocking issues when migrating logback configuration to log4j2 (with 
reference to [4] and [5])

  1. possibility to split the configuration in multiple files ([5] is 
actually part of [4])
  2. possibility to inject properties from external file(s) (see 
beginning of [5])
  3. possibility to declare a DataSource inside the JDBCAppender
  4. possibility to provide a custom connection source


>> Thanks for your support.
>> Regards.
>> [1] http://logging.apache.org/log4j/2.x/manual/configuration.html#AutomaticConfiguration

Francesco Chicchiriccò

ASF Member, Apache Syncope PMC chair, Apache Cocoon PMC Member

To unsubscribe, e-mail: log4j-user-unsubscribe@logging.apache.org
For additional commands, e-mail: log4j-user-help@logging.apache.org

View raw message