Page 1 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
Integrating Apache NiFi and Apache Flink
Feb 4th 2016
Bryan Bende – Member of Technical Staff
Page 2 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
Outline
• Introduction to NiFi
• NiFi Site-To-Site
• Flink + NiFi Integration
• Use Case Discussion
Page 3 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
About Me
• Member of Technical Staff at Hortonworks
• Apache NiFi Committer & PMC Member since June 2015
• Contributed NiFi + Flink Streaming Integration
• Twitter: @bbende / Blog: bryanbende.com
Page 4 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
Introduction to Apache NiFi
Page 5 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
Apache NiFi• Powerful and reliable system to process and
distribute data
• Directed graphs of data routing and transformation
• Web-based User Interface for creating, monitoring, & controlling data flows
• Highly configurable - modify data flow at runtime, dynamically prioritize data
• Data Provenance tracks data through entire system
• Easily extensible through development of custom components
[1] https://nifi.apache.org/
Page 6 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
NiFi - TerminologyFlowFile
• Unit of data moving through the system• Content + Attributes (key/value pairs)
Processor• Performs the work, can access FlowFiles
Connection• Links between processors• Queues that can be dynamically prioritized
Process Group• Set of processors and their connections• Receive data via input ports, send data via output ports
Page 7 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
NiFi - User Interface
• Drag and drop processors to build a flow• Start, stop, and configure components in real time• View errors and corresponding error messages• View statistics and health of data flow• Create templates of common processor & connections
Page 8 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
NiFi - Provenance
• Tracks data at each point as it flows through the system
• Records, indexes, and makes events available for display
• Handles fan-in/fan-out, i.e. merging and splitting data
• View attributes and content at given points in time
Page 9 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
NiFi - Queue Prioritization
• Configure a prioritizer per connection
• Determine what is important for your data – time based, arrival order, importance of a data set
• Funnel many connections down to a single connection to prioritize across data sets
• Develop your own prioritizer if needed
Page 10 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
NiFi - Extensibility
Built from the ground up with extensions in mind
Service-loader pattern for…• Processors• Controller Services• Reporting Tasks• Prioritizers
Extensions packaged as NiFi Archives (NARs)• Deploy NiFi lib directory and restart• Provides ClassLoader isolation• Same model as standard components
Page 11 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
NiFi - Architecture
OS/Host
JVM
Flow Controller
Web Server
Processor 1 Extension N
FlowFileRepository
ContentRepository
ProvenanceRepository
Local Storage
OS/Host
JVM
Flow Controller
Web Server
Processor 1 Extension N
FlowFileRepository
ContentRepository
ProvenanceRepository
Local Storage
OS/Host
JVM
Flow Controller
Web Server
Processor 1 Extension N
FlowFileRepository
ContentRepository
ProvenanceRepository
Local Storage
OS/Host
JVM
NiFi Cluster Manager – Request Replicator
Web Server
MasterNiFi Cluster Manager (NCM)
OS/Host
JVM
Flow Controller
Web Server
Processor 1 Extension N
FlowFileRepository
ContentRepository
ProvenanceRepository
Local Storage
SlavesNiFi Nodes
Page 12 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
NiFi Site-To-Site
Page 13 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
NiFi Site-To-Site
• Direct communication between two NiFi instances
• Push to Input Port on receiver, or Pull from Output Port on source
• Communicate between clusters, standalone instances, or both
• Handles load balancing and reliable delivery
• Secure connections using certificates (optional)
Page 14 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
Site-To-Site Push
• Source connects Remote Process Group to Input Port on destination
• Site-To-Site takes care of load balancing across the nodes in the cluster
NCM
Node 1
Input Port
Node 2
Input Port
Standalone NiFi
RPG
Page 15 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
Site-To-Site Pull
• Destination connects Remote Process Group to Output Port on the source
• If source was a cluster, each node would pull from each node in cluster
NCM
Node 1
RPG
Node 2
RPG
Standalone NiFi
Output Port
Page 16 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
Site-To-Site Client
• Code for Site-To-Site broken out into reusable module• https://github.com/apache/nifi/tree/master/nifi-commons/nifi-site-to-site-client
• Can be used from any Java program to push/pull from NiFi
Java Program
Site-To-Site Client
Node 1
Output Port
NCM
Node 2
Output Port
Page 17 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
Flink + NiFi Integration
Page 18 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
Flink + NiFi Integration
• Use Site-To-Site Client in Flink Streaming
• NiFiSource to pull data from NiFi Output Port
• NiFiSink to push data to NiFi Input Port
• NiFiDataPacket to represent data to/from NiFi (think FlowFile)
public interface NiFiDataPacket {
byte[] getContent();
Map<String, String> getAttributes();
}
Page 19 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
NiFi Source Example
StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
SiteToSiteClientConfig clientConfig = new SiteToSiteClient.Builder() .url("http://localhost:8080/nifi") .portName("Data for Flink") .requestBatchCount(…) .buildConfig();
SourceFunction<NiFiDataPacket> nifiSource = new NiFiSource(clientConfig);
DataStream<NiFiDataPacket> streamSource = env.addSource(nifiSource);
Page 20 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
NiFi Sink Example
StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
SiteToSiteClientConfig clientConfig = new SiteToSiteClient.Builder()
.url("http://localhost:8080/nifi") .portName("Data from Flink") .buildConfig();
// Creates a NiFiDataPacket from incoming data of a given type// Here we are creating NiFiDataPackets for each StringNiFiDataPacketBuilder<String> dpb = ...
DataStreamSink<String> dataStream = ... .addSink(new NiFiSink<>(clientConfig, dpb));
Page 21 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
Use Case Discussion
Page 22 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
Drive Data to Flink for Analysis
NiFi Flink
NiFi
NiFi
• Drive data from sources to central data center for analysis
• Tiered collection approach at various locations, think regional data centers
Edge
Edge
Core
Page 23 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
Dynamically Adjusting Data Flow
• Push analytic results from Flink back to NiFi
• Push results back to edge locations/devices to change behavior
NiFi Flink
NiFi
NiFi
Edge
Edge
Core
Page 24 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
1. Logs filtered by level and sent from Edge -> Core
2. Flink produces new filter levels based on rate & sends back to core
3. Edge polls core for new filter levels & updates filtering
Example: Dynamic Log Collection
Core NiFiFlink
Edge NiFiLogs Logs
New Filters
Logs Output Log Input Log Output
Result Input Store Result
Service Fetch ResultPoll Service
Filter
New Filters
New Filters
Poll
Analytic
Page 25 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
Dynamic Log Collection – Edge NiFi
Page 26 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
Dynamic Log Collection – Core NiFi
Page 27 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
Dynamic Log Collection – Flink StreamingStreamExecutionEnvironment env = ...
SiteToSiteClientConfig clientConfig = getSourceConfig(props);DataStream<NiFiDataPacket> streamSource = env.addSource(new NiFiSource(clientConfig));
int windowMs = ...LogLevelFlatMap logLevelFlatMap = new LogLevelFlatMap(...);
DataStream<LogLevels> counts = streamSource.flatMap(logLevelFlatMap) .timeWindowAll(Time.of(windowSize, TimeUnit.MILLISECONDS)) .apply(new LogLevelWindowCounter());
double rate = ...SiteToSiteClientConfig sinkConfig = getSinkConfig(props);NiFiDataPacketBuilder<LogLevels> builder = new DictionaryBuilder(window, rate);
counts.addSink(new NiFiSink<>(sinkConfig, builder));
Page 28 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
Dynamic Log Collection – Full Flow
NiFi Flink
NiFi
NiFi
Edge
Edge
Core
Logs
Logs
Logs
New Filters
New Filters
New Filters
Page 29 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
Summary
• Use NiFi to drive data from sources to Flink
• Leverage Flink results to adjust your dataflows
Sources• [1] https://nifi.apache.org/
Resources• https://github.com/bbende/nifi-streaming-examples• https://github.com/apache/flink/tree/master/flink-examples/flink-examples-streaming• https://flink.apache.org/news/2015/02/09/streaming-example.html
Contact Info: • Email: [email protected]• Twitter: @bbende
Page 30 © Hortonworks Inc. 2011 – 2015. All Rights Reserved
Thank you
Top Related