Flume-ng configuration with an HDFS sink

18 Jun 2013

I’ve been playing around with flume-ng and its HDFS sink recently to try to understand how I can stream data into HDFS and work with it using Hadoop. The documentation for flume-ng is unfortunately lacking, so I’ve typed up some quick notes on how to configure and test the HDFS sink.

This document assumes that you have Hadoop installed and running locally, with flume-ng version 1.2.0 or above.

In this example, the name of our agent is just agent. First, let’s define a channel for agent named memory-channel of type memory.

# Define a memory channel on agent called memory-channel.
agent.channels.memory-channel.type = memory

Next, let’s configure a source for agent, called tail-source, which watches the system.log file. Let us also assign it to the memory-channel.

# Define a source on agent and connect to channel memory-channel.
agent.sources.tail-source.type = exec
agent.sources.tail-source.command = tail -F /var/log/system.log
agent.sources.tail-source.channels = memory-channel

Now, configure two sinks: logger and HDFS. Then, we specify the path to the name node for HDFS, pointing to the output path of where we want the files stored.

# Define a sink that outputs to logger.
agent.sinks.log-sink.channel = memory-channel
agent.sinks.log-sink.type = logger

# Define a sink that outputs to hdfs.
agent.sinks.hdfs-sink.channel = memory-channel
agent.sinks.hdfs-sink.type = hdfs
agent.sinks.hdfs-sink.hdfs.path = hdfs://localhost:54310/tmp/system.log/
agent.sinks.hdfs-sink.hdfs.fileType = DataStream

Then, we configure the agent’s channels, sources and sinks.

# Finally, activate.
agent.channels = memory-channel
agent.sources = tail-source
agent.sinks = log-sink hdfs-sink

Finally, let’s start the flume agent, logging all output to the console, and starting agent agent.

# Run flume-ng, with log messages to the console.

$ bin/flume-ng agent --conf ./conf/ -f conf/flume.conf \

    -Dflume.root.logger=DEBUG,console -n agent

Finally, you should see output like this as data is written to the filesystem.

2013-06-18 14:00:49,784 (hdfs-hdfs-sink-call-runner-0) [INFO - org.apache.flume.sink.hdfs.BucketWriter.doOpen(BucketWriter.java:189)] Creating hdfs://localhost:54310/tmp/system.log//FlumeData.1371589249458.tmp