使用 Homebrew, chocolatey, MacPorts, Nix 安装 hadoop

项目链接

• 主页
• 仓库
• 软件包管理器页面

命令和别名

• FederationStateStore
• distribute-exclude.sh
• hadoop
• hadoop-daemon.sh
• hadoop-daemons.sh
• hdfs
• httpfs.sh
• kms.sh
• mapred
• mr-jobhistory-daemon.sh
• refresh-namenodes.sh
• start-all.sh
• start-balancer.sh
• start-dfs.sh
• start-secure-dns.sh
• start-yarn.sh
• stop-all.sh
• stop-balancer.sh
• stop-dfs.sh
• stop-secure-dns.sh
• stop-yarn.sh
• workers.sh
• yarn-daemon.sh
• yarn-daemons.sh

项目历史

Apache's early Hadoop documentation records the project split from Nutch MapReduce and distributed filesystem code in 2006, with Hadoop code moved into its own source tree in February, project approval in March, and release 0.1.0 in April. The project grew around a practical abstraction: store data across many machines with HDFS and run distributed computation with MapReduce.

Hadoop 1.0.0 arrived in December 2011 after what Apache called six years of gestation. That release came from the 0.20-security line and included security, HBase-related HDFS append/sync work, WebHDFS, and performance improvements. The 1.x era established Hadoop as a deployable foundation for large-scale storage and batch processing.

The 2.x line changed Hadoop's architecture by introducing YARN, described in Apache release notes as NextGen MapReduce. YARN separated cluster resource management from the MapReduce programming model, letting Hadoop run a wider set of distributed applications while keeping HDFS as a common storage layer. Apache's 2.0.2-alpha notes also recorded YARN deployment on a 2000-node cluster at release time.

Hadoop 3.x continued the infrastructure evolution. The 3.0.0 documentation highlights erasure coding in HDFS for lower storage overhead than three-way replication, along with a broader set of HDFS, YARN, MapReduce, and operational features. Later Apache site news documents 3.4 and 3.5 lines as large maintenance and enhancement releases.

采用历史

Hadoop's adoption history is tied to the rise of commodity-cluster data processing. The Apache homepage describes a framework designed to scale from single servers to thousands of machines, with failure detection and handling in the application layer rather than relying only on hardware high availability.

The broader ecosystem made Hadoop more than one binary. Apache release notes and docs refer to HBase, Pig, Oozie, Hive, Sqoop, WebHDFS, and YARN applications as downstream or companion systems. For package managers, that ecosystem means Hadoop often arrives as a dependency, local test harness, or compatibility target even when production clusters are provisioned by other means.

使用方式

On a workstation, the package is commonly used for command-line tools such as `hadoop`, `hdfs`, `yarn`, and `mapred`, for local pseudo-distributed testing, for interacting with HDFS-compatible services, or for building software that expects Hadoop client libraries and scripts. The single-node setup docs are a common starting point because they let developers run Hadoop locally before touching a cluster.

Operational Hadoop usage is configuration-heavy. Apache cluster setup documentation names site-specific XML files under `etc/hadoop`, including core-site.xml, hdfs-site.xml, yarn-site.xml, and mapred-site.xml, plus environment scripts such as hadoop-env.sh and yarn-env.sh. The same docs instruct administrators to distribute configuration to HADOOP_CONF_DIR across machines.

为什么软件包爱好者会关心

Hadoop is a heavyweight formula by package-manager standards: a Java-centric distributed-systems stack with many scripts, XML configs, daemon roles, and ecosystem expectations. It matters to package nerds because installing it locally can satisfy client tooling, build-time assumptions, demos, and compatibility tests without standing up a full distribution.

It is also a naming and dependency landmark. Many big-data packages historically needed to say which Hadoop line, HDFS client, YARN API, or MapReduce compatibility level they targeted. That makes the package less about a single executable and more about a versioned contract with the surrounding data ecosystem.

时间线

• 2006: Hadoop code moved out of Nutch, the project was approved, and release 0.1.0 was published.
• 2011: Hadoop 1.0.0 was released from the 0.20-security code line.
• 2012: Hadoop 2.0.0-alpha introduced HDFS HA, YARN, HDFS Federation, and protobuf wire compatibility.
• 2012: Hadoop 2.0.2-alpha notes recorded a more stable YARN deployed on a 2000-node cluster.
• 2017: Hadoop 3.0.0 documentation described HDFS erasure coding and other 3.x architecture updates.
• 2024: Apache announced the Hadoop 3.4 line with thousands of fixes, improvements, and enhancements since 3.3.
• 2026: Apache announced the Hadoop 3.5 line with hundreds of fixes, improvements, and enhancements since 3.4.

Related projects

• Apache Nutch: origin project from which Hadoop's early MapReduce and distributed filesystem code split.
• HDFS: Hadoop Distributed File System storage layer.
• YARN: Hadoop cluster resource-management framework.
• MapReduce: batch-processing framework and command surface in Hadoop.
• Apache HBase, Hive, Pig, Oozie, and Sqoop: common Hadoop-era ecosystem projects referenced in Apache release and setup material.

来源

• https://hadoop.apache.org/
• https://hadoop.apache.org/docs/r3.0.0/
• https://hadoop.apache.org/docs/stable/hadoop-project-dist/hadoop-common/ClusterSetup.html
• https://hadoop.apache.org/docs/stable/hadoop-project-dist/hadoop-common/SingleCluster.html
• https://hadoop.apache.org/release/2.0.0-alpha.html
• https://hadoop.apache.org/release/page/9.html
• https://hadoop.apache.org/version_control.html
• https://web.mit.edu/~mriap/hadoop/hadoop-0.13.1/docs/