TouK Nussknacker – using Apache Flink made easier for analysts and business

Few weeks ago we (TouK) revealed on Github our latest open source project – Nussknacker. What is it and why should you care?

Why?

First, some history: more than year ago one of our clients decided it’s high time for Real Time Marketing. They had pretty large data streams and lots of ideas for marketing campaigns, so one of the key success factors was the ability to process the data fast. We prepared a POC based on Apache Flink and it turned out that it’s really great piece of technology – fast and accurate.

There was just one problem – how to write and customize processes? Flink, as many modern stream processing engines, provides rich and friendly DSL, both in Java and Scala. But for our client that was not really enough. See, they are enterprise that do not employ developers – they have many decent, competent analysts who know the business and SQL – but not Java or (Heaven forbid…) Scala.

Nussknacker to the rescue!

So we decided to create a simple process designer for them. It looks more or less like this:

You draw a diagram, fill in the details (like filter expressions – “#input.usageData > 0” or SMS/mail content), then you press the Deploy button and voilà – your brand new process in running on Flink cluster!

Of course first somebody (that is, developer) has to prepare data sources (most probably Kafka topics), design data model (POJOs or case classes) and implement actions – like sending emails or sending some events to another Kafka topic. But once a model of data and external services are defined, an analyst can define and deploy processes all by him/herself.

More features

Sounds a bit scary to let your users run stream processes with GUI? Bad filter conditions can have serious performance implications if you deal with streams of tens of thousands of events per second… That’s why we let user test their diagrams first – each test case can be generated by sample of real data coming from e.g. Kafka and then run in Flink sandbox mini-cluster.

We have also many more features that make working with Nussknacker and Flink easier: subprocesses, versioning, generating PDF documentation, migration between environments and last but certainly not least – integration with InfluxDB/Grafana to provide detailed insight into how process is doing:

Where can I use it?

What are Nussknacker use cases? Our main deployments deal with RTM (Real Time Marketing). One of our clients started with RTM and then found out that Nussknacker is also great choice for fraud detection in real time. Industries we are working with include telcos, banks and media companies. We are also thinking about other possibilities – for example IoT.

Sounds cool?

If you are interested in easy access for semi-technical analysts to streaming data – give Nussknacker a try!

You can find the code at Github: https://github.com/touk/nussknacker, we also have a nice, Docker-based quickstart: https://touk.github.io/nussknacker/Quickstart.html.

And if you are coming to Flink Forward next week in Berlin – join me on Wednesday afternoon to hear more – .

In next days/weeks we’ll post more information on TouK blog both on Nussknacker architecture and internals and on interesting use cases – stay tuned :)

You May Also Like

Multi module Gradle project with IDE support

This article is a short how-to about multi-module project setup with usage of the Gradle automation build tool.

Here's how Rich Seller, a StackOverflow user, describes Gradle:
Gradle promises to hit the sweet spot between Ant and Maven. It uses Ivy's approach for dependency resolution. It allows for convention over configuration but also includes Ant tasks as first class citizens. It also wisely allows you to use existing Maven/Ivy repositories.
So why would one use yet another JVM build tool such as Gradle? The answer is simple: to avoid frustration involved by Ant or Maven.

Short story

I was fooling around with some fresh proof of concept and needed a build tool. I'm pretty familiar with Maven so created project from an artifact, and opened the build file, pom.xml for further tuning.
I had been using Grails with its own build system (similar to Gradle, btw) already for some time up then, so after quite a time without Maven, I looked on the pom.xml and found it to be really repulsive.

Once again I felt clearly: XML is not for humans.

After quick googling I found Gradle. It was still in beta (0.8 version) back then, but it's configured with Groovy DSL and that's what a human likes :)

Where are we

In the time Ant can be met but among IT guerrillas, Maven is still on top and couple of others like for example Ivy conquer for the best position, Gradle smoothly went into its mature age. It's now available in 1.3 version, released at 20th of November 2012. I'm glad to recommend it to anyone looking for relief from XML configured tools, or for anyone just looking for simple, elastic and powerful build tool.

Lets build

I have already written about basic project structure so I skip this one, reminding only the basic project structure:
<project root>

├── build.gradle
└── src
├── main
│ ├── java
│ └── groovy

└── test
├── java
└── groovy
Have I just referred myself for the 1st time? Achievement unlocked! ;)

Gradle as most build tools is run from a command line with parameters. The main parameter for Gradle is a 'task name', for example we can run a command: gradle build.
There is no 'create project' task, so the directory structure has to be created by hand. This isn't a hassle though.
Java and groovy sub-folders aren't always mandatory. They depend on what compile plugin is used.

Parent project

Consider an example project 'the-app' of three modules, let say:
  1. database communication layer
  2. domain model and services layer
  3. web presentation layer
Our project directory tree will look like:
the-app

├── dao-layer
│ └── src

├── domain-model
│ └── src

├── web-frontend
│ └── src

├── build.gradle
└── settings.gradle
the-app itself has no src sub-folder as its purpose is only to contain sub-projects and build configuration. If needed it could've been provided with own src though.

To glue modules we need to fill settings.gradle file under the-app directory with a single line of content specifying module names:
include 'dao-layer', 'domain-model', 'web-frontend'
Now the gradle projects command can be executed to obtain such a result:
:projects

------------------------------------------------------------
Root project
------------------------------------------------------------

Root project 'the-app'
+--- Project ':dao-layer'
+--- Project ':domain-model'
\--- Project ':web-frontend'
...so we know that Gradle noticed the modules. However gradle build command won't run successful yet because build.gradle file is still empty.

Sub project

As in Maven we can create separate build config file per each module. Let say we starting from DAO layer.
Thus we create a new file the-app/dao-layer/build.gradle with a line of basic build info (notice the new build.gradle was created under sub-project directory):
apply plugin: 'java'
This single line of config for any of modules is enough to execute gradle build command under the-app directory with following result:
:dao-layer:compileJava
:dao-layer:processResources UP-TO-DATE
:dao-layer:classes
:dao-layer:jar
:dao-layer:assemble
:dao-layer:compileTestJava UP-TO-DATE
:dao-layer:processTestResources UP-TO-DATE
:dao-layer:testClasses UP-TO-DATE
:dao-layer:test
:dao-layer:check
:dao-layer:build

BUILD SUCCESSFUL

Total time: 3.256 secs
To use Groovy plugin slightly more configuration is needed:
apply plugin: 'groovy'

repositories {
mavenLocal()
mavenCentral()
}

dependencies {
groovy 'org.codehaus.groovy:groovy-all:2.0.5'
}
At lines 3 to 6 Maven repositories are set. At line 9 dependency with groovy library version is specified. Of course plugin as 'java', 'groovy' and many more can be mixed each other.

If we have settings.gradle file and a build.gradle file for each module, there is no need for parent the-app/build.gradle file at all. Sure that's true but we can go another, better way.

One file to rule them all

Instead of creating many build.gradle config files, one per each module, we can use only the parent's one and make it a bit more juicy. So let us move the the-app/dao-layer/build.gradle a level up to the-app/build-gradle and fill it with new statements to achieve full project configuration:
def langLevel = 1.7

allprojects {

apply plugin: 'idea'

group = 'com.tamashumi'
version = '0.1'
}

subprojects {

apply plugin: 'groovy'

sourceCompatibility = langLevel
targetCompatibility = langLevel

repositories {
mavenLocal()
mavenCentral()
}

dependencies {
groovy 'org.codehaus.groovy:groovy-all:2.0.5'
testCompile 'org.spockframework:spock-core:0.7-groovy-2.0'
}
}

project(':dao-layer') {

dependencies {
compile 'org.hibernate:hibernate-core:4.1.7.Final'
}
}

project(':domain-model') {

dependencies {
compile project(':dao-layer')
}
}

project(':web-frontend') {

apply plugin: 'war'

dependencies {
compile project(':domain-model')
compile 'org.springframework:spring-webmvc:3.1.2.RELEASE'
}
}

idea {
project {
jdkName = langLevel
languageLevel = langLevel
}
}
At the beginning simple variable langLevel is declared. It's worth knowing that we can use almost any Groovy code inside build.gradle file, statements like for example if conditions, for/while loops, closures, switch-case, etc... Quite an advantage over inflexible XML, isn't it?

Next the allProjects block. Any configuration placed in it will influence - what a surprise - all projects, so the parent itself and sub-projects (modules). Inside of the block we have the IDE (Intellij Idea) plugin applied which I wrote more about in previous article (look under "IDE Integration" heading). Enough to say that with this plugin applied here, command gradle idea will generate Idea's project files with modules structure and dependencies. This works really well and plugins for other IDEs are available too.
Remaining two lines at this block define group and version for the project, similar as this is done by Maven.

After that subProjects block appears. It's related to all modules but not the parent project. So here the Groovy language plugin is applied, as all modules are assumed to be written in Groovy.
Below source and target language level are set.
After that come references to standard Maven repositories.
At the end of the block dependencies to groovy version and test library - Spock framework.

Following blocks, project(':module-name'), are responsible for each module configuration. They may be omitted unless allProjects or subProjects configure what's necessary for a specific module. In the example per module configuration goes as follow:
  • Dao-layer module has dependency to an ORM library - Hibernate
  • Domain-model module relies on dao-layer as a dependency. Keyword project is used here again for a reference to other module.
  • Web-frontend applies 'war' plugin which build this module into java web archive. Besides it referes to domain-model module and also use Spring MVC framework dependency.

At the end in idea block is basic info for IDE plugin. Those are parameters corresponding to the Idea's project general settings visible on the following screen shot.


jdkName should match the IDE's SDK name otherwise it has to be set manually under IDE on each Idea's project files (re)generation with gradle idea command.

Is that it?

In the matter of simplicity - yes. That's enough to automate modular application build with custom configuration per module. Not a rocket science, huh? Think about Maven's XML. It would take more effort to setup the same and still achieve less expressible configuration quite far from user-friendly.

Check the online user guide for a lot of configuration possibilities or better download Gradle and see the sample projects.
As a tasty bait take a look for this short choice of available plugins:
  • java
  • groovy
  • scala
  • cpp
  • eclipse
  • netbeans
  • ida
  • maven
  • osgi
  • war
  • ear
  • sonar
  • project-report
  • signing
and more, 3rd party plugins...