For systems consisting of thousands and possibly even tens of thousands of
classes,
are there scalability advantages to naming loggers by their package name
only,
as opposed to scoping completely down to the class name? (Specifically
referring to
the Logger.getLogger(a.b.class); call.)
I'm
Howdy,
We have systems with thousands of classes, each having a private static
final Logger with its name as the full class name. We haven't run into
scalability or performance problems related to this.
Losing the ability to turn the debug level for a specific class (as
opposed to a package) at
Mike -
I use class level granularity (as recommended by several members of this
list) and have not seen performance or resource issues arise using 100s (if
not 1000+) of loggers. Personally I have not had much need for class-level
control of my loggers, but we still are in the habit of removing
Sorry, my intent wasn't to recommend a change in the naming convention in
general or start controversy.
I simply don't know a lot about the resources used per logger, and am just
looking for guidance/advice on how to name loggers in large systems.
It appears your advice is to include the class
Howdy,
Sorry, my intent wasn't to recommend a change in the naming convention
in
general or start controversy.
Don't worry, I didn't interpret it as such ;) Just explaining some of
our (and by our I mean the projects I work on, not the log4j developer
community) design and experiences in this
To get around the problems I was having with log4j in a clustered
environment (multiple processes on one filesystem), I wrote a new appender
that writes to the servlet log as if you called servletContext.log(...).
Has anyone thought of doing this before? Hadn't seen much on the lists
about it. I
Hello Bill,
Sounds interesting. Can you post it along with instructions on use?
I'd like to try it.
Jake
Friday, January 17, 2003, 12:30:32 PM, you wrote:
BS To get around the problems I was having with log4j in a clustered
BS environment (multiple processes on one filesystem), I wrote a new
I have a webapp that uses log4j to direct logs to several files. Within
the tomcat installation used for development the logs go where
expected. However in the tomcat installation to be used for production
each log entry goes to the expected file and an additional entry is
written to
At 11:06 17.01.2003 -0800, you wrote:
There are so many places to look trying to track this down involving
tomcat and commons-logging, so I don't quite know exactly what to ask of
this list. I don't think there are any competing configurations, so I
don't think a different appender with an
Here is the story.
The commons-logging API, when it decides that you want to use log4j,
takes the liberty of configuring a ConsoleAppender (on System.Err) and
attaching it to the root logger. The PattenLayout uses the pattern %r
[%t] %p %c{2} %x - %m%n. See the code of
10 matches
Mail list logo