The Oracle Big Data SQL installation automatically copies the OHSH installation kit to each node where Oracle Big Data SQL is installed. There are some manual steps you must perform to complete the installation.

1. Log on to the Oracle Database server as the oracle user.

2. Set up OHSH to work with Copy to Hadoop.

   1. Create a staging directory in the file system on the Oracle Database node:

      $ mkdir <OS_STAGE_DIR>

   2. Create an Oracle directory on this staging directory and grant read/write privileges to the Oracle Database user (oracle or other). In this example, ORACLE_STAGE_DIR refers to the name of the directory object in Oracle Database. OS_STAGE_DIR refers to the path on the server.

      SQL> CREATE DIRECTORY <ORACLE_STAGE_DIR> AS '<OS_STAGE_DIR>';
      SQL> GRANT READ,WRITE ON DIRECTORY <ORACLE_STAGE_DIR> TO <ORACLE_USER>;

      Note:

      Later, if you employ the “stage” method to copy Oracle tables, use the OHSH set locationdirectory command to point to the Oracle directory object name (the name of the directory you created above in this step).

   3. If you are connecting to a CDH-based Hadoop system (Oracle Big Data Appliance or other), perform the substeps below to enable the OHSH “FUSE” method, in which you copy files to Hive tables over an HDFS Mountable FUSE directory. Note that HDP does not support Fuse, therefore you cannot create a mount point to HDFS on HDP using these steps.

      1. On the Oracle Database host, create a FUSE mount to the root of the HDFS system on the Hadoop cluster. Refer to Cloudera’s documentation on installing FUSE. You may find current instructions at Configuring Mountable HDFS on Cloudera’s website.

      2. Create the Oracle directory on the FUSE mount and grant privileges. In this example, ORACLE_FUSE_DIR refers to the name of the directory object in Oracle Database. OS_FUSE_MOUNT refers to the path to the mount point on the server.

               SQL> CREATE DIRECTORY <ORACLE_FUSE_DIR> AS '<OS_FUSE_MOUNT>';
               SQL> GRANT READ,WRITE ON DIRECTORY <ORACLE_FUSE_DIR> TO <ORACLE_USER>;

      Note that when you use OHSH later to copy Oracle tables to Hive using the "FUSE" method, use the OHSH set locationdirectory command to point to the same <ORACLE_FUSE_DIR>

3. Edit $ORACLE_HOME/ohsh/bin/ohsh_config.sh in order to configure the home directories of dependencies. When OHSH is installed by Oracle Big Data SQL, most of the required variables are already set for you. Set HS2_HOST_PORT and (if needed) HIVE_SESS_VAR_LIST as follows.

   export HS2_HOST_PORT=<HS2_HOST>:<HS2_PORT>
   

   Also set HIVE_SESS_VAR_LIST to any required Hive session variables. For example:

   • To connect to HiveServer2 with Kerberos authentication: export HIVE_SESS_VAR_LIST="principal=<The server principal of HiveServer2>"

     The Hive principal  is specified by the configuration property hive.server2.authentication.kerberos.principal in hive-site.xml.

   • To connect to HiveServer2 running in HTTP mode: export HIVE_SESS_VAR_LIST="transportMode=http;httpPath=<The HTTP endpoint>"

   • To connect to HiveServer2 when SSL is enabled: export HIVE_SESS_VAR_LIST="ssl=true;sslTrustStore=<Trust Store path>;trustStorePassword=<Trust Store password>"

4. If you want to configure OHSH to use Oracle Wallet, set environment the variables TNS_ADMIN and WALLET_LOCATION to the directory path where TNS and Oracle Wallet are stored.

   export WALLET_LOCATION="<wallet_location>"
   export TNS_ADMIN="<tns_admin_location>"

   Note:

   Later, when you start an OHSH session, if the TNS_ADMIN and WALLET_LOCATION are different on the Hadoop cluster then set hadooptnsadmin and hadoopwalletlocation:

   set hadooptnsadmin <tns_admin_hadoop_node>
   set hadoopwalletlocation <wallet_location_hadoop_node>

5. Clear the HADOOP_CLASSPATH environment variable before invoking Oracle Shell for Hadoop Loaders.

6. Start an OHSH session and check for resources. This check indicates whether or not the configuration is correct.

   $ ohsh
   ohsh> show resources

   You should see the resources hadoop0, hive0, and bash0.

On the Hadoop cluster management server, Oracle Big Data SQL installs OHSH from an RPM to /opt/oracle/ohsh. Perform the steps below as root in order to complete the Hadoop side of the configuration.

1. If the Hadoop system is not an Oracle Big Data Appliance, then download the SQL*Plus Instant Client and Oracle Instant Client for Linux from the Oracle Technology Network. Select the client versions that matches the version of the Oracle Database.

   Extract both packages into the same directory.

   Add the directory to PATH and LD_LIBRARY_PATH.

2. Create a local directory for TNS and from TNS_ADMIN on the Oracle Database host, copy the tnsnames.ora and sqlnet.ora file into this directory.

   If you create an Oracle Wallet for the Oracle Database host, then also copy over the Oracle Wallet file from the database server. Also check that sqlnet.ora points to the correct Oracle Wallet location.

   WALLET_LOCATION= 
         (SOURCE=(METHOD=FILE)(METHOD_DATA= 
           (DIRECTORY=<your wallet directory>))) 
      sqlnet.wallet_override=true

3. As root, edit /opt/oracle/ohsh/bin/ohsh_config.sh. Set the following environment variables.

   export CP2HADOOP_HOME=/opt/oracle/orahivedp-<version 
   export HS2_HOST_PORT=<HiveServer2 hostname>:<Port number>
   

   Also set HIVE_SESS_VAR_LIST to any required Hive session variables. For example:

   • To connect to HiveServer2 with Kerberos authentication: export HIVE_SESS_VAR_LIST="principal=<The server principal of HiveServer2>"

   • To connect to HiveServer2 running in HTTP mode: export HIVE_SESS_VAR_LIST="transportMode=http;httpPath=<The HTTP endpoint>"

   • To connect to HiveServer2 when SSL is enabled: export HIVE_SESS_VAR_LIST="ssl=true;sslTrustStore=<Trust Store path>;trustStorePassword=<Trust Store password>"

4. If TNS and Oracle Wallet are enabled, also add the following variables to the file and export them:

   export WALLET_LOCATION="<your wallet directory>"
   export TNS_ADMIN="<your TNS directory>"

5. Add /opt/oracle/ohsh and $HADOOP_HOME/bin to the PATH variable

   PATH=/opt/oracle/ohsh-<version>/bin:$PATH
   PATH=$HADOOP_HOME/bin:$PATH 

6. Clear the HADOOP_CLASSPATH environment variable before invoking Oracle Shell for Hadoop Loaders.

7. Start an OHSH session and check for resources. This check indicates whether or not the configuration is correct.

   $ ohsh
   ohsh> show resources

   You should see the resources hadoop0, hive0, and bash0.

See $ORACLE_HOME/ohsh/doc and $ORACLE_HOME/ohsh/examples for examples that show how to run the OHSH load methods.

See the Oracle Big Data SQL User’s Guide more information on Copy to Hadoop and Oracle Shell for Hadoop Loaders.

When OHSH jobs are executed, status and other information about the job is recorded into a back-end database. To access information from the OHSH command line, you must first configure the connection to the database.