[logback-dev] [JIRA] Created: (LBCLASSIC-243) Scribe Appender

[Gordon Sims (JIRA)]

Scribe Appender
---------------

                 Key: LBCLASSIC-243
                 URL: http://jira.qos.ch/browse/LBCLASSIC-243
             Project: logback-classic
          Issue Type: New Feature
          Components: appender
            Reporter: Gordon Sims
            Assignee: Logback dev list


(posting on behalf of colleague Derek Libby)

We have written a ScribeAppender for internal use at ITA Software which we would like to give back to the Logback project.  The starting point for this appender was the log4j ScribeAppender documented at:

   https://github.com/pnayak/Scribe-logback-Appender/blob/master/src/java/org/apache/log4j/scribe/ScribeAppender.java

The appender currently accepts the following configuration settings:

   host - The host name on which the Scribe server is running

   port - The port on which the Scribe server is listening

   category - The category name to use when publishing to Scribe

   bucketSize - The number of messages to accumulate before publishing to Scribe

   connectInterval - The number of milliseconds to wait between reconnect attempts when connection is lost.  The appender is blocked during this time

Issues:

   The most obvious issue to me is the new dependencies this appender introduces on the scribe, thrift and fb303 jars.  To my knowledge these are not available in any of the public maven repositories.  Details on how I built these dependences in house are attached to this mail.  Because they are not readily available adding them as dependencies to logback-classic is probably to burdensome.  Perhaps creating a logback-extras or even a logback-scribe jar for people to use if they so choose is the way to go.

   The fact that logback encoders require a stream to which to output the results of their formatting was a bit of a pain point.  I worked around this by passing a ByteArrayOutputStream to the appenders configured encoder and then extracting the results from it after then encode step is completed.  I also have to check for and remove newlines from encoded log lines as there is no need for them when publishing messges to Scribe.

We have been using this encoder on a couple of data center machines for about a month now with no ill effects.

--

Instructions for building the jars necessary for Java to communicate with a scribe server:

Install Thrift:
    Download the thrift tarball from http://thrift.apache.org/download/
    tar -xzf thrift-0.5.0.tar.gz
    # Order seemed to matter with the next two lines
    sudo apt-get install libboost-test1.40-dev libevent-dev automake libtool flex bison pkg-config g++
    sudo apt-get install libboost-dev
    sudo apt-get install php5-dev
    sudo apt-get install python-dev
    cd $THRIFT_HOME
    ./configure
    make
    sudo make install
    
Build thrift jars
    cd $THRIFT_HOME/lib/java
    ant
    cd $THRIFT_HOME/contrib/fb303/java
    ant
    
Build scribe classes
    cd $SCRIBE_ROOT/if
    # Add:
    #    include "/path/to/your/thrift/contrib/fb303/if/fb303.thrift"
    # to $SCRIBE_HOME/if/scribe.thrift
    thrift --gen java scribe.thrift
    


-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: http://jira.qos.ch/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira