Hari Bhutanadhu — Published On March 7, 2023 and Last Modified On March 8th, 2023
Beginner Data Engineering Database Hadoop


Apache Flume is a tool/service/data ingestion mechanism for gathering, aggregating, and delivering huge amounts of streaming data from diverse sources, such as log files, events, and so on, to centralized data storage. Flume is a tool that is very dependable, distributed, and customizable. einsteinerupload of.

In this article, we will discuss about Apache Flume, Flume Architecture, installation, setup, and Configuration. Basically, data from multiple sources can be transferred to centralized storage or processing systems like HDFS, HBase, and Spark using the Flume platform, a distributed, highly reliable, and scalable platform. Applications that process and analyze big data use Flume in the Apache Hadoop ecosystem.

Apache flume
Source: Analytics Vidhya

Learning Objectives

The learning objectives in this article are

  • Understanding the architecture of flume.
  • Flume installation and configuration.
  • Sources of flume sink of the flume and channels of the flume.
  • Draw light on improving the performance of the flume.

This article was published as a part of the Data Science Blogathon.

Table of Contents

Apache Flume Architecture

The Flume architecture comprises three main components: sources, channels, and sinks.

  • Sources: Sources are the components in the Flume architecture that collect data from various sources, such as log files, network sockets, or system processes. Flume provides several built-in source types, including exec, spooling directory, and netcat. Custom sources can also be developed to support ingestion from various data sources. The data collected by the sources is typically unstructured and needs to be processed and transformed before it can be stored or analyzed.
  • Channels: Channels are the intermediate buffers that store and transport data between sources and sinks. Flume provides several built-in channel types: memory, JDBC, and file. Channels can be configured to have different properties, such as the maximum size of events or the maximum number of events stored. The data stored in channels can be compressed and encrypted, and it is configurable to be durable or not.
  • Sinks: Sinks are the final destinations that receive and process data from the channels, such as HDFS, HBase, Elasticsearch, or Kafka. Flume provides several built-in sink types, including HDFS, HBase, Solr, and Elasticsearch. Custom sinks can also be developed to support delivery to various data destinations.
  • Agents: Agents are the logical nodes in the Flume architecture. They are responsible for receiving data from sources, storing it in channels, and delivering it to sinks. An agent can have one or more sources, channels, and sinks. Flume agents can be configured to run on multiple nodes in a distributed manner to provide scalability and fault tolerance.
  • Event: Event is the basic unit of data in the Flume architecture. An event is a piece of unstructured data that is generated by a source and transported through channels to a sink. Events can have headers and bodies, which interceptors can modify and filter.
  • Interceptor: Interceptor is a component in the Flume architecture that can modify or filter events as they pass through the data flow. Interceptors can be used to enrich events with additional metadata or filter out unwanted events.

Apache Flume Installation and Configuration

Steps for installing and configuring Flume

1. Download Flume: Download the latest version of Flume from the Apache Flume website.

2. Install Java: Flume requires Java to be installed on the system. Make sure Java is installed and configured properly.

3. Set up environment variables: Set up environment variables for Java and Flume in the system. For example, add the following lines to the .bashrc or .bash_profile file:

export JAVA_HOME=/path/to/java
export PATH=$PATH:$JAVA_HOME/bin
export FLUME_HOME=/path/to/flume

4. Configure Flume: Flume is configured using configuration files written in the properties format. Create a configuration file called flume.conf in the $FLUME_HOME/conf directory. Here’s an example configuration for a simple Flume agent that reads data from a file and writes it to the console:

agent.sources = mysource
agent.channels = mychannel
agent.sinks = mysink

agent.sources.mysource.type = exec
agent.sources.mysource.command = tail -F /var/log/syslog

agent.sinks.mysink.type = logger

agent.channels.mychannel.type = memory
agent.channels.mychannel.capacity = 1000
agent.channels.mychannel.transactionCapacity = 100

agent.sources.mysource.channels = mychannel
agent.sinks.mysink.channel = mychannel

This configuration defines a source called “mysource” that reads data from a syslog file using the “exec” type source, a sink called “mysink” that logs data to the console using the “logger” type sink, and a memory channel called “mychannel” that stores events in memory. The source and sink are connected to the channel using the channels property.

5. Start Flume: Start Flume by running the following command in the terminal:

$ flume-ng agent -n agent -c $FLUME_HOME/conf -f $FLUME_HOME/conf/flume.conf

This command starts a Flume agent with the name “agent”, using the configuration file located in the $FLUME_HOME/conf directory.

6. Verify Flume is running: Verify that Flume is running by checking the log files in the $FLUME_HOME/logs directory.

Apache Flume Sources

  • Avro Source: This source receives data over the network via the Apache Avro protocol. It can be used to collect data from applications written in any programming language.
  • Netcat Source: This source listens on a network socket and receives data as plain text. It is useful for collecting data from sources that can send data as plain text over the network.
  • Syslog Source: This source receives data from Syslog servers. It is useful for collecting data from network devices, Unix servers, and other Syslog-enabled devices.
  • Spooling Directory Source: This source monitors a directory for new files and reads them as they are added to the directory. It is useful for collecting data from sources that write data to files.
  • HTTP Source: This source receives data over HTTP. It is useful for collecting data from web servers, REST APIs, and other HTTP-enabled sources.
  • JMS Source: This source receives data from a JMS (Java Messaging Service) queue. It is useful for collecting data from applications that send data to a JMS queue.

Apache Flume Sinks

  • HDFS Sink: This sink delivers the data to the Hadoop Distributed File System (HDFS). It is useful for storing data in Hadoop for further processing and analysis.
  • Kafka Sink: This sink delivers the data to the Apache Kafka messaging system. It is useful for processing data in real-time using Kafka.
  • ElasticSearch Sink: This sink delivers the data to the ElasticSearch search engine. It is useful for indexing and searching the data in real-time.
  • File Roll Sink: This sink delivers the data to files on the local file system. It is useful for storing data in a local file system for backup and archival purposes.
  • Hive Sink: This sink delivers the data to Apache Hive, a data warehousing and SQL-like query language for Hadoop. It is useful for processing and analyzing data using Hive.
  • JDBC Sink: This sink delivers the data to a relational database using JDBC. It is useful for storing data in a relational database for further processing and analysis.

Flume Channels

Flume channels can be of two types: memory-based channels and file-based channels.

  • Memory-based channels: Memory-based channels store events in memory, which makes them faster than file-based channels. They are useful when the volume of data being processed is small enough to fit into memory.
  • File-based channels: File-based channels store events on disk, which makes them slower than memory-based channels but allows them to handle large volumes of data that cannot fit into memory.

Flume channels have a configurable capacity, determining the maximum number of events that can be stored in the channel at any given time. If the channel becomes full, new events will be dropped until space becomes available. To avoid this, it is important to appropriately configure the channel capacity based on the data volume being processed.

Flume channels also support various transactional mechanisms to ensure that events are delivered to the sink in a reliable and consistent manner. For example, file-based channels use a write-ahead log (WAL) to ensure that events are not lost in case of a system failure.

 Source: flume.apache.org
Source: flume.apache.org

Flume Performance Tuning

To achieve optimal performance, it is important to tune Flume based on the data collection and delivery task requirements.

  • Use the appropriate sources and sinks: Choose the appropriate sources and sinks based on the data collection and delivery task requirements. Using the right source and sink can significantly improve performance.
  • Use multiple channels: Flume channels buffer and store the events that are being processed by the sources and sinks. Using multiple channels can increase the parallelism of the Flume pipeline and improve performance.
  • Increase the channel capacity: By default, Flume channels have a small capacity. Increasing the channel capacity can improve the performance of the Flume pipeline by reducing the number of events that are dropped due to buffer overflow.
  • Use memory-based channels: Flume provides both disk-based and memory-based channels. Memory-based channels are faster than disk-based channels and are recommended for high-performance environments.
  • Use compression: Flume provides built-in support for compression. Enabling compression can reduce the size of the data being transferred and improve the performance of the Flume pipeline.
  • Tune the batch size: Flume processes events in batches. Tuning the batch size based on the data collection and delivery task requirements can improve performance.
  • Use load balancing: If multiple Flume agents are processing data, use load balancing to distribute the workload evenly across the agents. This can improve the overall performance of the Flume pipeline.

Application of Flume

Flume can be used to gather information from Internet of Things (IoT) devices, such as sensors or smart home devices, and store it in a centralized database.

Flume can be used to gather information from websites such as Twitter and Facebook and then store it in a centralized database for analysis.

These are some of the applications of Flume. This data is useful for the company’s future analysis and upcoming market strategies.

Advantages of Apache Flume

Many advantages of Apache Flume include scalable, reliability, and centralization; It has Large Data sets, It has Real-time data streaming, and many more.

  • Open-Source: Apache Flume is an open-source distributed system. So it is available free of cost.
  • Inexpensive: It is less costly to install and operate. Its maintenance is very economical.
  • Scalable: It is designed to handle massive amounts of data in a distributed manner. It can be set up to operate across several servers in a group, allowing it to handle data increases amounts without losing performance.
  • Data Ingestion: Data integrity is protected by built-in methods in Flume like error checking and data validation, which help ensure that data is not lost or changed during transmission.
  • Fault Tolerance: It has the ability to find the error. It contains techniques for moving information over many nodes in the cluster and for each unsuccessful operation and making sure that data is not lost due to any failure.
  • Reliable: These above features make Apache Flume a reliable tool for data.

Click here for more details

apache flume
Source: Edureka

Disadvantages of Apache Flume

  • Limited Scaling: Apache Flume scalability is frequently low because it can be difficult for any company to size the hardware of a standard Apache Flume, and it is typically a trial-and-error process. As a result, the scalability of Flume is frequently challenged.
  • Duplicacy: Apache Flume does not provide a guarantee that all the messages received will be completely unique. It is possible for duplicate messages to appear in certain situations.
  • Reliability Issues Sometimes: If the backing store is not chosen wisely, then there may be scalability and reliability issues.

These are some of the disadvantages of Flume


In conclusion, Flume is a good data ingestion and processing tool that can handle massive amounts of data from various sources and connect with Hadoop and other big data technologies without issues. Users may scale as required, adapt data flows, and set up data pipelines quickly. Flume’s robust data management features to ensure data security, availability, and availability, and its connection with other tools allows users to do advanced analyses and gain valuable info. Ultimately, Flume can help companies manage and get insights from their data to promote innovation and economic progress.

  • We can store the data in any of the centralized storage using Apache Flume (HBase, HDFS).
  • When the rate of incoming data exceeds the rate at which it can be written to the destination, Flume operates as a bridge between data producers and centralized stores, ensuring a continual flow of data.
  • Flume includes the capability of contextual routing.
  • Flume transactions are channel-based, with each communication requiring two transactions (one sender and one receiver). It ensures consistent message transmission.
  • Flume is a dependable, fault-tolerant, scalable, managed, and adaptable application.

The media shown in this article is not owned by Analytics Vidhya and is used at the Author’s discretion.

About the Author

Our Top Authors

Download Analytics Vidhya App for the Latest blog/Article