This document details the internals of how the sqoop-server works.

This document is relevant to the release 1.99.5. Further changes can happen in future releases

Sqoop Tomcat Server

  • Sqoop-server uses tomcat web server, it is very bare bones.
  • The main entry point is the TomcatToolRunner, it bootstraps the tomcat and loads all the sqoop related classes into its class path.  It is invoked from the bash script .

    /sqoop.sh server start 
  • The main hook for the sqoop server to start is this entry in the web.xml. Tomcat invokes it callbacks as it bootstraps and we use the contextInitialized callback to initialize all the related code.

 

<!-- Listeners -->
   <listener>
    <listener-class>org.apache.sqoop.server.ServerInitializer</listener-class>
  </listener>
 

 

Sqoop Server

  • The sqoop server is represented by the java class SqoopServer.java
  • SqoopServer.initialize() is the called from the ServerInitiaizer
  • SqoopServer.destroy() is called when the tomcat server is shutdown

Sqoop Servlets

  <!-- Version servlet -->
  <servlet>
    <servlet-name>VersionServlet</servlet-name>
    <servlet-class>org.apache.sqoop.server.VersionServlet</servlet-class>
    <load-on-startup>1</load-on-startup>
  </servlet>
  <servlet-mapping>
    <servlet-name>VersionServlet</servlet-name>
    <url-pattern>/version</url-pattern>
  </servlet-mapping>
   <!-- Generic Configurable servlet -->
  <servlet>
    <servlet-name>v1.ConfigurableServlet</servlet-name>
    <servlet-class>org.apache.sqoop.server.v1.ConfigurableServlet</servlet-class>
    <load-on-startup>1</load-on-startup>
  </servlet>
  <servlet-mapping>
    <servlet-name>v1.ConfigurableServlet</servlet-name>
    <url-pattern>/v1/configurable/*</url-pattern>
  </servlet-mapping>
  <!-- Connector servlet -->
  <servlet>
    <servlet-name>v1.ConnectorServlet</servlet-name>
    <servlet-class>org.apache.sqoop.server.v1.ConnectorServlet</servlet-class>
    <load-on-startup>1</load-on-startup>
  </servlet>
  <servlet-mapping>
    <servlet-name>v1.ConnectorServlet</servlet-name>
    <url-pattern>/v1/connector/*</url-pattern>
  </servlet-mapping>
  <!-- Connectors servlet -->
  <servlet>
    <servlet-name>v1.ConnectorsServlet</servlet-name>
    <servlet-class>org.apache.sqoop.server.v1.ConnectorServlet</servlet-class>
    <load-on-startup>1</load-on-startup>
  </servlet>
  <servlet-mapping>
    <servlet-name>v1.ConnectorsServlet</servlet-name>
    <url-pattern>/v1/connectors/*</url-pattern>
  </servlet-mapping>
  <!-- Driver servlet -->
  <servlet>
    <servlet-name>v1.DriverServlet</servlet-name>
    <servlet-class>org.apache.sqoop.server.v1.DriverServlet</servlet-class>
    <load-on-startup>1</load-on-startup>
  </servlet>
  ......
  <!-- Job servlet -->
  <servlet>
    <servlet-name>v1.JobServlet</servlet-name>
    <servlet-class>org.apache.sqoop.server.v1.JobServlet</servlet-class>
    <load-on-startup>1</load-on-startup>
  </servlet>
  <servlet-mapping>
    <servlet-name>v1.JobServlet</servlet-name>
    <url-pattern>/v1/job/*</url-pattern>
  </servlet-mapping>
  <!-- Jobs servlet -->
  <servlet>
    <servlet-name>v1.JobsServlet</servlet-name>
    <servlet-class>org.apache.sqoop.server.v1.JobsServlet</servlet-class>
    <load-on-startup>1</load-on-startup>
  </servlet>
  <servlet-mapping>
    <servlet-name>v1.JobsServlet</servlet-name>
    <url-pattern>/v1/jobs/*</url-pattern>
  </servlet-mapping>
  <!-- Submissions servlet -->
  <servlet>
    <servlet-name>v1.SubmissionsServlet</servlet-name>
    <servlet-class>org.apache.sqoop.server.v1.SubmissionsServlet</servlet-class>
    <load-on-startup>1</load-on-startup>
  </servlet>
  <servlet-mapping>
    <servlet-name>v1.SubmissionsServlet</servlet-name>
    <url-pattern>/v1/submissions/*</url-pattern>
  </servlet-mapping>

</web-app>

 

  • There is an authentication filter to authenticate all request.

 

<!-- Filter -->
<filter>
  <filter-name>authFilter</filter-name>
  <filter-class>org.apache.sqoop.filter.SqoopAuthenticationFilter</filter-class>
</filter>

 

  • There are two authentication mode supported: simple and Kerberos, which could be set in the sqoop.properties.

 

#
# Authentication configuration
#
org.apache.sqoop.security.authentication.type=SIMPLE
org.apache.sqoop.security.authentication.handler=org.apache.sqoop.security.Authentication.SimpleAuthenticationHandler
org.apache.sqoop.security.authentication.anonymous=true

#org.apache.sqoop.security.authentication.type=KERBEROS
#org.apache.sqoop.security.authentication.handler=org.apache.sqoop.security.Authentication.KerberosAuthenticationHandler
#org.apache.sqoop.security.authentication.kerberos.principal=sqoop/_HOST@NOVALOCAL
#org.apache.sqoop.security.authentication.kerberos.keytab=/home/kerberos/sqoop.keytab
#org.apache.sqoop.security.authentication.kerberos.http.principal=HTTP/_HOST@NOVALOCAL
#org.apache.sqoop.security.authentication.kerberos.http.keytab=/home/kerberos/sqoop.keytab
#org.apache.sqoop.security.authentication.enable.doAs=true
#org.apache.sqoop.security.authentication.proxyuser.#USER#.users=*
#org.apache.sqoop.security.authentication.proxyuser.#USER#.groups=*
#org.apache.sqoop.security.authentication.proxyuser.#USER#.hosts=*

 

Sqoop Request Handlers

 

Each Sqoop Servlet has its corresponding handler class that handles the request for that servlet. It then internally calls the internal sqoop core/ common code.

public interface RequestHandler {
  static final String CONNECTOR_NAME_QUERY_PARAM = "cname";
  static final String JOB_NAME_QUERY_PARAM = "jname";
  JsonBean handleEvent(RequestContext ctx);
}
public class ConnectorRequestHandler implements RequestHandler {
...
}

Sqoop Client

  • Sqoop Client is represented by the java class SqoopClient.java
  • It has wrapper ResourceRequest classes for each sqoop entity, they encapsulate the request/postBody parameters to be sent in the request. Refer to Sqoop 2 (1.99.4) Entity Nomenclature and Relationships for more details on the supported Sqoop Entities.
  • It used the bare bones HttpURLConnection object to make requests to the Sqoop-server.

    HttpURLConnection conn = new DelegationTokenAuthenticatedURL().openConnection(url, authToken);

     SqoopClient used to use jersey REST client for making tomcat requests. Recently it was switched to Hadoop-auth/SPENGO for adding Kerberos support that are documented here

    https://cwiki.apache.org/confluence/display/SQOOP/Security+Guide+On+Sqoop+2

  • Run command to start Sqoop Client.

 

/sqoop.sh client

 

  • In Kerberos Authentication mode. Kinit is required to set Kerberos environment.

 

kinit sqoop/server-fqdn@HADOOP.COM

 

 

  • No labels