如何启动Datanode？ （找不到 start-dfs.sh 脚本）

Question

我们正在无头系统上设置自动部署：因此这里不能选择使用 GUI。

Hortonworks Data Platform 中 hdfs 的 start-dfs.sh 脚本在哪里？ CDH/cloudera 将这些文件打包在 hadoop/sbin 目录下。然而，当我们在 HDP 下搜索这些脚本时，却找不到它们：

$ pwd
/usr/hdp/current

HDP 中存在哪些脚本？

[stack@s1-639016 current]$ find -L . -name \*.sh
./hadoop-hdfs-client/sbin/refresh-namenodes.sh
./hadoop-hdfs-client/sbin/distribute-exclude.sh
./hadoop-hdfs-datanode/sbin/refresh-namenodes.sh
./hadoop-hdfs-datanode/sbin/distribute-exclude.sh
./hadoop-hdfs-nfs3/sbin/refresh-namenodes.sh
./hadoop-hdfs-nfs3/sbin/distribute-exclude.sh
./hadoop-hdfs-secondarynamenode/sbin/refresh-namenodes.sh
./hadoop-hdfs-secondarynamenode/sbin/distribute-exclude.sh
./hadoop-hdfs-namenode/sbin/refresh-namenodes.sh
./hadoop-hdfs-namenode/sbin/distribute-exclude.sh
./hadoop-hdfs-journalnode/sbin/refresh-namenodes.sh
./hadoop-hdfs-journalnode/sbin/distribute-exclude.sh
./hadoop-hdfs-portmap/sbin/refresh-namenodes.sh
./hadoop-hdfs-portmap/sbin/distribute-exclude.sh
./hadoop-client/sbin/hadoop-daemon.sh
./hadoop-client/sbin/slaves.sh
./hadoop-client/sbin/hadoop-daemons.sh
./hadoop-client/etc/hadoop/hadoop-env.sh
./hadoop-client/etc/hadoop/kms-env.sh
./hadoop-client/etc/hadoop/mapred-env.sh
./hadoop-client/conf/hadoop-env.sh
./hadoop-client/conf/kms-env.sh
./hadoop-client/conf/mapred-env.sh
./hadoop-client/libexec/kms-config.sh
./hadoop-client/libexec/init-hdfs.sh
./hadoop-client/libexec/hadoop-layout.sh
./hadoop-client/libexec/hadoop-config.sh
./hadoop-client/libexec/hdfs-config.sh
./zookeeper-client/conf/zookeeper-env.sh
./zookeeper-client/bin/zkCli.sh
./zookeeper-client/bin/zkCleanup.sh
./zookeeper-client/bin/zkServer-initialize.sh
./zookeeper-client/bin/zkEnv.sh
./zookeeper-client/bin/zkServer.sh

注意：有零启动/停止 sh 脚本..

我对启动 namenode(s)、journalnode 和 datanodes 的 start-dfs.sh 脚本特别感兴趣。

Answer 1

如何启动DataNode

su - hdfs -c "/usr/lib/hadoop/bin/hadoop-daemon.sh --config /etc/hadoop/conf start datanode";

Github - Hortonworks Start Scripts

更新

决定自己去找。

1. 使用 Ambari 启动单个节点，安装 HDP 2.2 (a)、HDP 2.3 (b)
2. sudo find / -name \*.sh | grep start

3. 找到

   (a)/usr/hdp/2.2.8.0-3150/hadoop/src/hadoop-hdfs-project/hadoop-hdfs/src/main/bin/s‌​tart-dfs.sh

   奇怪的是它在/usr/hdp/current 中不存在，应该是符号链接的。

   (b)/hadoop/yarn/local/filecache/10/mapreduce.tar.gz/hadoop/sbin/start-dfs.sh

Answer 2

管理您的 hadoop 集群的推荐方法是通过管理员面板。由于您正在开发 Hotronworks 发行版，因此使用 Ambari 更有意义。