How To setup Hadoop To Stream Apache Logs to HBase Via Flume

This repository contains sample code to configure Flume to stream Apache web logs to HBase on Hadoop. It includes components such as:

1. The Flume configuratione

2. The Apache Configuration

3. A custom Java implementation of AsyncHbaseEventSerializer

4. A rotate log script

5. DDL the HBase table

Assumes you have working knowledge of these technical components.

Step 1: Deploy HBase Table

1. Confirm hbase-site.xml configuration

   1. vi /etc/hbase/conf/hbase-site.xml

   2. Confirm rootdir has proper hostname

   3. Repeat for

      a. Core-site.xml

      b. Hive-site.xml

2. Make sure HBase services started

   a) service hbase-master stop/start

   b) service hbase-regionserver stop/start

3. Create HBase table

   a) sudo -u hdfs hbase shell

     For help "help<enter>"
   

   Ex: help 'create'

   b) create 'apache_access_log', {NAME => 'common'}, {NAME => 'http'}, {NAME => 'misc'}

Step 2: Apache Log File format and rotation

1. Install apache

   a) yum install httpd

   b) vi /etc/httpd/conf/httpd.conf

   c) Add the following at the end to create custom log that is easily processed / parseable

   Thanks to http://www.cepheid.org/~jeff/?p=30

   logs for Flume injestion. Custom log format for easier regex matching

   LogFormat "%{UNIQUE_ID}e %v %h %l %u %{%Y%m%d%H%M%S}t %r %>s %b %{Referer}i %{User-Agent}i" vfcombined CustomLog "| /usr/sbin/rotatelogs /var/log/httpd/al_flume.%Y%m%d%H%M%S 60" vfcombined

   d) Uncomment

   a. #LoadModule unique_id_module modules/mod_unique_id.so
   

   d) chkconfig httpd on

   e) /etc/init.d/httpd restart

2. Create Log Rotation Script

See http://www.cepheid.org/~jeff/?p=30 for original script

a) mkdir /home/cloudera/logs

b) mkdir /home/cloudera/logs/flumelog

c) Create script flume_rotate.sh in /home/cloudera/logs, make sure permissions are X for root

   	#!/bin/sh
   	# current time minus 61 seconds (latest file we want to move)
   	# used as a filename and timestamp for a file (used for a find(8) comparison)
   	# Thanks to http://www.cepheid.org/~jeff/?p=30
   	STAMP=`date --date="61 seconds ago" "+%Y%m%d%H%M.%S"`
   	# where rotatelogs is writing
   	SDIR=/var/log/httpd
   	# where flume is picking up
   	DDIR=/var/log/httpd/flumelog
   	# debugging
   	# echo "Date: " `date`
   	# echo "Using: $STAMP"
   	# Create our comparison file
   	touch -t $STAMP $SDIR/$STAMP
   	# move older logs to flume spooldir
   	find $SDIR -name ‘al_flume.*’ ‘!’ -newer $SDIR/$STAMP -exec mv "{}" $DDIR/ \;
   	# cleanup touchfile
   	rm $SDIR/$STAMP
   

3. Schedule every 60 seconds in cron

   a. crontab -e

    * * * * * /home/cloudera/logs/flume_rotate.sh
   

Step 3: Create Custom Event Serializer

This custom Java class takes the log event from the serialized Flume source and maps each log field to an HBase column family

See https://blogs.apache.org/flume/entry/streaming_data_into_apache_hbase for original SplitteSerializer, which was modified for our purposes

1. Create Java Class

   a) cd /home/cloudera/logs/

   b) mkdir com; mkdir com/hbase; mkdir com/hbase/log; mkdir com/hbase/log/util

   c) vi com/hbase/log/util/AsyncHbaseLogSerializer.java

   d) see Java code under com/hbase/log/util of this repository, paste contents

2. Compile Java Class

   a) export CLASSPATH=/usr/lib/flume-ng/lib/flume-ng-core-1.3.0.jar:/usr/lib/flume-ng/lib/flume-ng-sdk-1.3.0.jar:/usr/lib/flume-ng/lib/flume-thrift-source-1.2.0-cdh4.1.3.jar:/usr/lib/flume-ng/lib/flume-scribe-source-1.2.0-cdh4.1.3.jar:/usr/lib/flume-ng/lib/flume-hdfs-sink-1.2.0-cdh4.1.3.jar:/usr/lib/flume-ng/lib/flume-ng-node-1.2.0-cdh4.1.3.jarflume-ng-log4jappender-1.2.0-cdh4.1.3.jar:/usr/lib/flume-ng/lib/flume-avro-source-1.2.0-cdh4.1.3.jar:/usr/lib/flume-ng/lib/flume-irc-sink-1.2.0-cdh4.1.3.jar:/usr/lib/flume-ng/lib/flume-ng-configuration-1.2.0-cdh4.1.3.jar:/usr/lib/flume-ng/lib/flume-file-channel-1.2.0-cdh4.1.3.jar:/usr/lib/hadoop/flume-sources-1.0-SNAPSHOT.jar:/usr/lib/flume-ng/lib/asynchbase-1.2.0.jar:/usr/lib/hive/lib/hbase.jar:/usr/lib/flume-ng/lib/flume-ng-hbase-sink-1.2.0-cdh4.1.3.jar:/usr/lib/flume-ng/lib/guava-11.0.2.jar

   b) javac com/hbase/log/util/AsyncHbaseLogEventSerializer.java

   c) jar cf LogEventUtil.jar com

   d) jar tf LogEventUtil.jar com

   e) chmod 775 LogEventUtil.jar

   f) cp LogEventUtil.jar /usr/lib/flume-ng/lib

Step 4: Configure Flume

1. See flume-conf/flume.conf for proper values

2. Make sure you have flume core and flume ng sdk 1.3.0 or greater

3. Also need to download and add the folloiwng missing jars

    cp flume-ng-core-1.3.0.jar /usr/lib/flume-ng/lib/
    
    cp /home/cloudera/facebook/apache-flume-1.3.0-bin/lib/flume-ng-sdk-1.3.0.jar /usr/lib/flume-ng/lib
   

4. /etc/init.d/flume-ng-agent stop

5. /etc/init.d/flume-ng-agent start

6. make sure everything looks good in /var/log/flume-ng/flume.log