Sqoop ships as one binary package that incorporates two separate parts - client and server.
- Server You need to install server on single node in your cluster. This node will then serve as an entry point for all Sqoop clients.
- Client Clients can be installed on any number of machines.
1.1.1. Server installation¶
Copy the Sqoop artifact to the machine where you want to run Sqoop server. The Sqoop server acts as a Hadoop client, therefore Hadoop libraries (Yarn, Mapreduce, and HDFS jar files) and configuration files (core-site.xml, mapreduce-site.xml, ...) must be available on this node. You do not need to run any Hadoop related services - running the server on a “gateway” node is perfectly fine.
You should be able to list a HDFS for example:
hadoop dfs -ls
Sqoop currently supports Hadoop version 2.6.0 or later. To install the Sqoop server, decompress the tarball (in a location of your choosing) and set the newly created forder as your working directory.
# Decompress Sqoop distribution tarball tar -xvf sqoop-<version>-bin-hadoop<hadoop-version>.tar.gz # Move decompressed content to any location mv sqoop-<version>-bin-hadoop<hadoop version>.tar.gz /usr/lib/sqoop # Change working directory cd /usr/lib/sqoop
184.108.40.206. Hadoop dependencies¶
Sqoop server needs following environmental variables pointing at Hadoop libraries - $HADOOP_COMMON_HOME, $HADOOP_HDFS_HOME, $HADOOP_MAPRED_HOME and $HADOOP_YARN_HOME. You have to make sure that those variables are defined and pointing to a valid Hadoop installation. Sqoop server will not start if Hadoop libraries can’t be found.
The Sqoop server uses environment variables to find Hadoop libraries. If the environment variable $HADOOP_HOME is set, Sqoop will look for jars in the following locations: $HADOOP_HOME/share/hadoop/common, $HADOOP_HOME/share/hadoop/hdfs, $HADOOP_HOME/share/hadoop/mapreduce and $HADOOP_HOME/share/hadoop/yarn. You can specify where the Sqoop server should look for the common, hdfs, mapreduce, and yarn jars indepently with the $HADOOP_COMMON_HOME, $HADOOP_HDFS_HOME, $HADOOP_MAPRED_HOME and $HADOOP_YARN_HOME environment variables.
# Export HADOOP_HOME variable export HADOOP_HOME=/... # Or alternatively HADOOP_*_HOME variables export HADOOP_COMMON_HOME=/... export HADOOP_HDFS_HOME=/... export HADOOP_MAPRED_HOME=/... export HADOOP_YARN_HOME=/...
If the environment $HADOOP_HOME is set, Sqoop will usee the following locations: $HADOOP_HOME/share/hadoop/common, $HADOOP_HOME/share/hadoop/hdfs, $HADOOP_HOME/share/hadoop/mapreduce and $HADOOP_HOME/share/hadoop/yarn.
220.127.116.11. Hadoop configuration¶
Sqoop server will need to impersonate users to access HDFS and other resources in or outside of the cluster as the user who started given job rather then user who is running the server. You need to configure Hadoop to explicitly allow this impersonation via so called proxyuser system. You need to create two properties in core-site.xml file - hadoop.proxyuser.$SERVER_USER.hosts and hadoop.proxyuser.$SERVER_USER.groups where $SERVER_USER is the user who will be running Sqoop 2 server. In most scenarios configuring * is sufficient. Please refer to Hadoop documentation for details how to use those properties.
Example fragment that needs to be present in core-site.xml file for case when server is running under sqoop2 user:
<property> <name>hadoop.proxyuser.sqoop2.hosts</name> <value>*</value> </property> <property> <name>hadoop.proxyuser.sqoop2.groups</name> <value>*</value> </property>
If you’re running Sqoop 2 server under a so called system user (user with ID less then min.user.id - 1000 by default), then YARN will by default refuse to run Sqoop 2 jobs. You will need to add the user name who is running Sqoop 2 server (most likely user sqoop2) to a allowed.system.users property of container-executor.cfg. Please refer to YARN documentation for further details.
Example fragment that needs to be present in container-executor.cfg file for case when server is running under sqoop2 user:
18.104.22.168. Third party jars¶
To propagate any third party jars to Sqoop server classpath, create a directory anywhere on the file system and export it’s location in SQOOP_SERVER_EXTRA_LIB variable.
# Create directory for extra jars mkdir -p /var/lib/sqoop2/ # Copy all your JDBC drivers to this directory cp mysql-jdbc*.jar /var/lib/sqoop2/ cp postgresql-jdbc*.jar /var/lib/sqoop2/ # And finally export this directory to SQOOP_SERVER_EXTRA_LIB export SQOOP_SERVER_EXTRA_LIB=/var/lib/sqoop2/
Sqoop doesn’t ship with any JDBC drivers due to incompatible licenses. You will need to use this mechanism to install all JDBC drivers that are needed.
22.214.171.124. Configuring PATH¶
All user and administrator facing shell commands are stored in bin/ directory. It’s recommended to add this directory to your $PATH for easier execution, for example:
The remainder of the Sqoop 2 documentation assumes that the shell commands are in your $PATH.
126.96.36.199. Configuring Server¶
Server configuration files are stored in conf directory. File sqoop_bootstrap.properties specifies which configuration provider should be used for loading configuration for rest of Sqoop server. Default value PropertiesConfigurationProvider should be sufficient.
Second configuration file called sqoop.properties contains remaining configuration properties that can affect Sqoop server. The configuration file is very well documented, so check if all configuration properties fits your environment. Default or very little tweaking should be sufficient in most common cases.
188.8.131.52. Repository Initialization¶
The metadata repository needs to be initialized before starting Sqoop 2 server for the first time. Use Upgrade tool to initialize the repository:
You can verify if everything have been configured correctly using Verify tool:
sqoop2-tool verify ... Verification was successful. Tool class org.apache.sqoop.tools.tool.VerifyTool has finished correctly
184.108.40.206. Server Life Cycle¶
After installation and configuration you can start Sqoop server with following command:
You can stop the server using the following command:
By default Sqoop server daemon use port 12000. You can set org.apache.sqoop.jetty.port in configuration file conf/sqoop.properties to use different port.
1.1.2. Client installation¶
Just copy Sqoop distribution artifact on target machine and unzip it in desired location. You can start client with following command:
You can find more documentation for Sqoop shell in Command Line Shell.
Client is not acting as a Hadoop client and thus you do not need to be installed on node with Hadoop libraries and configuration files.