John
News http://wfbcivicfoundation.org/bv/w_news.php John
RE: configuring a non-core appender in log4j2
That did the trick. Thank you for your help. -Original Message- From: Zabicki Roman (HERE/Chicago) Sent: Friday, March 21, 2014 5:25 PM To: Log4J Users List Subject: RE: configuring a non-core appender in log4j2 when I build -- you've described the problem exactly. I'm building with maven, and we combine all of our classes and all the classes from the jars we're dependent on into one single jar. This single jar can, of course, only have 1 Log4j2Plugin.dat. I have to stick with this build/deployment strategy. I can't deploy multiple jars. So I will use the packages attribute and let you know on Monday how it goes. Thanks for pointing that out. I had read that page before, but I didn't get it. -Original Message- From: Ralph Goers [mailto:ralph.go...@dslextreme.com] Sent: Friday, March 21, 2014 3:56 PM To: Log4J Users List Subject: Re: configuring a non-core appender in log4j2 If the log4j flume jar is in the class path it should automatically be included as it has its own Log4j2Plugin.dat file. However, you can always manually cause plugins to be located by specifying the packages attribute - see http://logging.apache.org/log4j/2.x/manual/configuration.html#XML. Again though, if it is not automatically finding the Flume Appender then there is either something wrong in your configuration or the log4j-flume-ng jar is not in the classpath. I'm curious though - when you say when I build what exactly are you building? You cannot build a new jar that includes the contents of log4j-core.jar and the contents of other jars that also have plugins as each has their own Log4j2Plugin.dat and only one of them will wind up in your new jar. To do that, yes you would have to run the PluginManager against the contents of your new jar to generate your own custom Log4j2Plugin.dat. However, doing that is not recommended. We recommend that you leave the Log4j jars alone and include them in your classpath. Ralph On Mar 21, 2014, at 12:30 PM, Zabicki Roman (HERE/Chicago) roman.zabi...@here.com wrote: Hi, I want to use a non-core appender with log4j2. In particular, I want to use the apache flume appender. (http://logging.apache.org/log4j/2.x/log4j-flume-ng/) When I build, I wind up with the Log4j2Plugins.dat from log4j, which doesn't have support for the flume appender, so I can't register the flume appender in my log4j2.xml. It seems that there's no way to register custom appender classes just via log4j2.xml, and instead I have to use PluginManager to register the \org\apache\logging\log4j\core\config\plugins\Log4j2Plugin.dat. I can't find any examples of how to do this, let alone examples of how to automate this in a build. Is it possible to register a custom appender class solely through log4j2.xml, or do I need to go through the PluginManager? Whatever the answer, can someone show me an example? I'm particularly interested in how to automate this in a build. Thanks, Roman - To unsubscribe, e-mail: log4j-user-unsubscr...@logging.apache.org For additional commands, e-mail: log4j-user-h...@logging.apache.org
Running disruptor async performance tests
I am trying to run the async performance tests described on the link below on my machine. http://logging.apache.org/log4j/2.x/manual/async.html#Asynchronous_Logging_Performance I am not an expert with log4j, so here is how far I got after building log4j with maven: java -cp target/classes:target/test-classes:lib/disruptor-3.2.1.jar:../log4j-api/target/classes org.apache.logging.log4j.core.async.perftest.PerfTest org.apache.logging.log4j.core.async.perftest.RunLog4j2 blah blah.log 1 -verbose Then I get this in the output: avg=17 99%=32 99.99%=64 sampleCount=500 9962247 operations/second The questions I have are: 1. It looks like the source code IPerfTestRunner uses a much shorter message Short Msg instead of the 500 characters message stated in the link above. Is that intentional or is it a bug? Do we want to test the latency with the 500-character message or just a short message? 2. I notice that my logs are NOT going to any file. I am probably misconfiguring something with log4j. How do I generate a file with the messages from the performance test? 3. I just want to test with one asynchronous logging thread, so I am passing threadCount 1 above. What does the second parameter 'blah' mean? 4. Not sure why I get operations/seconds if I am not passing -throughput in the command-line. I just want to get the latency numbers for now. After that I will worry about throughput. So basically I just want to run the same test you run to see those great numbers on my production machine. Thanks for the help! -Becky
Re: Running disruptor async performance tests
Going to guess Remko has something useful to say on this, but since he's in Japan, just wait a bit for the timezones to make sense. On 28 March 2014 22:03, Rebecca Ahlvarsson rahlvars...@gmail.com wrote: I am trying to run the async performance tests described on the link below on my machine. http://logging.apache.org/log4j/2.x/manual/async.html#Asynchronous_Logging_Performance I am not an expert with log4j, so here is how far I got after building log4j with maven: java -cp target/classes:target/test-classes:lib/disruptor-3.2.1.jar:../log4j-api/target/classes org.apache.logging.log4j.core.async.perftest.PerfTest org.apache.logging.log4j.core.async.perftest.RunLog4j2 blah blah.log 1 -verbose Then I get this in the output: avg=17 99%=32 99.99%=64 sampleCount=500 9962247 operations/second The questions I have are: 1. It looks like the source code IPerfTestRunner uses a much shorter message Short Msg instead of the 500 characters message stated in the link above. Is that intentional or is it a bug? Do we want to test the latency with the 500-character message or just a short message? 2. I notice that my logs are NOT going to any file. I am probably misconfiguring something with log4j. How do I generate a file with the messages from the performance test? 3. I just want to test with one asynchronous logging thread, so I am passing threadCount 1 above. What does the second parameter 'blah' mean? 4. Not sure why I get operations/seconds if I am not passing -throughput in the command-line. I just want to get the latency numbers for now. After that I will worry about throughput. So basically I just want to run the same test you run to see those great numbers on my production machine. Thanks for the help! -Becky -- Matt Sicker boa...@gmail.com
