Skip to content

Latest commit

 

History

History
123 lines (85 loc) · 5.11 KB

README.md

File metadata and controls

123 lines (85 loc) · 5.11 KB
Raw

Flink Streaming with Kakfa

During this XKE we want to experiment with Flink a distributed processing engine like Spark. Flink exists of a Streaming and a DataSet-API. We are going to look into the Streaming-API.

What you need

Install the required tools which can be accomplised by downloading the distributions and unzip. For Mac users 'brew' might be handy.

  • Java 8
  • Kafka
  • Flink
  • Zookeeper (but thats bundled with Kafka and Flink)
  • Maven3 for building the fat-jar

Create a Project

Create a quickstart java Maven project based on the flink-quickstart archetype.

# this will create a subdir 'flink-streaming'
mvn archetype:generate\
    -DarchetypeGroupId=org.apache.flink\
    -DarchetypeArtifactId=flink-quickstart-java\
    -DarchetypeVersion=1.1.1\
    -DgroupId=com.gdd.tryouts\
    -DartifactId=flink-streaming\
    -Dversion=0.1\
    -Dpackage=com.gdd\
    -DinteractiveMode=false

And dependency an extra dependency for Kafka in the pom.xml:

	<dependency>
    <groupId>org.apache.flink</groupId>
    <artifactId>flink-connector-kafka-0.8_2.10</artifactId>
    <version>${flink.version}</version>
</dependency>

Build it with 'mvn package'.

Start a Flink

Start Flink and open the webinterface:

FLINK_HOME=$(pwd)
./bin/start-local.sh

Kafka the pub-sub message bus:

Start Kafka and create a topic:

# Start zookeeper & start Kafka
bin/zookeeper-server-start.sh -daemon config/zookeeper.properties
bin/kafka-server-start.sh config/server.properties

# Create a Kafka topic:
bin/kafka-topics.sh --list --zookeeper localhost:2181
bin/kafka-topics.sh --create --zookeeper localhost:2181 --replication-factor 1 --partitions 1 --topic texttopic
# Look there it is:
bin/kafka-topics.sh --list --zookeeper localhost:2181

How it Works?!

You have to submit your application (fat-jar) to Flink, so first build your jar with maven package. Go to the WebUI or use $FLINK_HOME/bin/flink run target/flink-streaming-0.1.jar [program arguments]. Note: in the pom.xml the mainClass is defined line ~350!

Exercise 1: Count words

Send some text into the Kafka-topic with the Console-Producer and Parse these streaming events with Flink. Print the top words.

bin/kafka-console-producer.sh --broker-list localhost:9092 --topic texttopic
$FLINK_HOME/bin/flink run target/flink-streaming-0.1.jar --kafkabroker localhost:9092 --kafka.topic texttopic --kafka.groupId myGroup1
# Or if you have commented out the mainClass in the pom.xml
$FLINK_HOME/bin/flink run target/flink-streaming-0.1.jar --class com.gdd.solutions.KafkaStreamingWordCount --kafkabroker localhost:9092 --kafka.topic texttopic --kafka.groupId myGroup1

Exercise 2: Sliding Window - WordCount

One of Flinks' strong features is the windows functions. These are not limited to ingest event times! Try make your streaming wordcount more fancy with Trending Words:

  • Window over the last minute
  • Let the window slide over 30 seconds
  • Print only the top 1 words (every minute)

Exercise 3: Twitter treding topics!

If you have completed the previous example you are very close of implementing a Treding Twitter Topics application! You need a Twitter api-token and consumer-key to slurp tweets from the interwebz. You can add the flink-twitter-source to create your trending topics.

Change your mainClass in the pom.xml to TwitterStreamProcessing and build the jar
$FLINK_HOME/bin/flink run target/flink-streaming-0.1.jar --token <YourToken> --token.secret <YourTokenSecret> --consumer.key <YourKey> --consumer.secret <YourConsuerSecret>
  • Extract the #tags from the tweets
  • Print the top 3 most tweeted hashtags over 5 minutes
  • Check if your results match :-)

Loading the HeroesOfTheStorm-replay dataset in Kafka

Instead of the console-producer we are going to load some data

  • Replays (1 row per match): ReplayID (Unique per match), Game Mode (3=Quick Match, 4=Hero League, 5=Team League), Map, Replay Length, Timestamp (UTC)

  • Replay Characters (10 rows per match, one for each Hero): ReplayID (Unique per match, links to other file), Is Auto Select, Hero, Hero Level, Is Winner, MMR Before

    #Download and unzip (brew install p7zip wget) wget https://d1i1jxrdh2kvwy.cloudfront.net/Data/HOTSLogs%20Data%202015-05-14%20-%202015-05-24.zip 7z x HOTSLogs*.zip

    $KAFKA_HOME/bin/kafka-console-producer.sh --topic hotsReplays --broker-list localhost:9092 < Replays*.csv $KAFKA_HOME/bin/kafka-console-producer.sh --topic hotsCharacters --broker-list localhost:9092 < ReplayCharacters*.csv

Troubleshoot:

Caused by: java.lang.NoClassDefFoundError: scala/collection/GenTraversableOnce$class Build jar with -Pbuild-jar to create FAT jar with deps.

Caused by: java.lang.RuntimeException: Unable to retrieve any partitions for the requested topics [texttopic].Please check previous log entries Topic does not exists!

Pom.xml: <mainClass>com.gdd.KafkaStreamingWordCount</mainClass>