Requirements for HDFS

From HVR
Jump to: navigation, search

This section describes the requirements, access privileges, and other features of HVR when using Hadoop Distributed File System (HDFS) for replication. For information about compatibility and supported versions of HDFS with HVR platforms, see Platform Compatibility Matrix.

To quickly setup replication into HDFS, see Quick Start for HVR into HDFS or Quick Start for HVR into HDFS with Hive.

HDFS
Capture Hub Integrate
Icon-Yes.png Icon-No.png Icon-Yes.png

Location Connection

This section lists and describes the connection details required for creating HDFS location in HVR.

SC-Hvr-Location Hive HDFS dec02.png
Field Description
Database Connection
Namenode The hostname of the HDFS NameNode.
  Example: myHDFSnode
Port The port on which the HDFS server (Namenode) is expecting connections.
  Example: 8020
Login The username to connect HVR to the HDFS Namenode.
  Example: hvruser
Credentials The credential (Kerberos Ticket Cache file) for the Login to connect HVR to the HDFS Namenode.
This field should be left blank to use a keytab file for authentication or if Kerberos is not used on the hadoop cluster. For more details, see HDFS Authentication and Kerberos.
Directory The directory path in the HDFS Namenode to be used for replication.
  Example: /user/hvr/


Hive ODBC Connection

HVR allows you to create Hive External Tables above HDFS which are only used during compare. You can enable/disable the Hive configuration for HDFS in location creation screen using the field Hive External Tables.

Field Description
Hive ODBC Connection
Hive Server Type The type of Hive server.  Available options:
  • Hive Server 1 (default): The driver connects to a Hive Server 1 instance.
  • Hive Server 2: The driver connects to a Hive Server 2 instance.
Service Discovery Mode The mode for connecting to Hive. This field is enabled only if Hive Server Type is Hive Server 2.  Available options:
  • No Service Discovery (default): The driver connects to Hive server without using the ZooKeeper service.
  • ZooKeeper: The driver discovers Hive Server 2 services using the ZooKeeper service.
Host(s) The IP address or hostname of the Hive server.
When Service Discovery Mode is ZooKeeper, specify the list of ZooKeeper servers in following format [ZK_Host1]:[ZK_Port1],[ZK_Host2]:[ZK_Port2], where [ZK_Host] is the IP address or hostname of the ZooKeeper server and [ZK_Port] is the TCP port that the ZooKeeper server uses to listen for client connections.
  Example: hdp24
Port The TCP port that the Hive server uses to listen for client connections. This field is enabled only if Service Discovery Mode is No Service Discovery.
  Example: 1000
Database The name of the database schema to use when a schema is not explicitly specified in a query.
  Example: mytestdb
ZooKeeper Namespace The namespace on ZooKeeper under which Hive Server 2 nodes are added. This field is enabled only if Service Discovery Mode is ZooKeeper.
Authentication
Mechanism The authentication mechanism for connecting HVR to Hive Server 2.  This field is enabled only if Hive Server Type is Hive Server 2. Available options:
  • No Authentication (default)
  • User Name
  • User Name and Password
  • Kerberos
User The username to connect HVR to Hive server. This field is enabled only if Mechanism is User Name or User Name and Password.
  Example: hadoop
Password The password of the User to connect HVR to Hive server. This field is enabled only if Mechanism is User Name and Password.
Service Name The Kerberos service principal name of the Hive server. This field is enabled only if Mechanism is Kerberos.
Host The Fully Qualified Domain Name (FQDN) of the Hive Server 2 host. The value of Host can be set as _HOST to use the Hive server hostname as the domain name for Kerberos authentication.
If Service Discovery Mode is disabled, then the driver uses the value specified in the Host connection attribute.
If Service Discovery Mode is enabled, then the driver uses the Hive Server 2 host name returned by ZooKeeper.
This field is enabled only if Mechanism is Kerberos.
Realm The realm of the Hive Server 2 host.
It is not required to specify any value in this field if the realm of the Hive Server 2 host is defined as the default realm in Kerberos configuration. This field is enabled only if Mechanism is Kerberos.
Linux / Unix
Driver Manager Library The directory path where the Unix ODBC Driver Manager Library is installed. For a default installation, the ODBC Driver Manager Library is available at /usr/lib64 and does not need to specified. When UnixODBC is installed in for example /opt/unixodbc-2.3.2 this would be /opt/unixodbc-2.3.2/lib.
ODBCSYSINI The directory path where odbc.ini and odbcinst.ini files are located. For a default installation, these files are available at /etc and does not need to be specified. When UnixODBC is installed in for example /opt/unixodbc-2.3.2 this would be /opt/unixodbc-2.3.2/etc.
ODBC Driver The user defined (installed) ODBC driver to connect HVR to the Hive server.
SSL Options Displays the SSL options.

SSL Options

SC-Hvr-Location Hive SSL.png
Field Description
Enable SSL Enable/disable (one way) SSL. If enabled, HVR authenticates the Hive server by validating the SSL certificate shared by the Hive server.
Two-way SSL Enable/disable two way SSL. If enabled, both HVR and Hive server authenticate each other by validating each others SSL certificate. This field is enabled only if Enable SSL is selected.
SSL Public Certificate The directory path where the .pem file containing the client's SSL public certificate is located. This field is enabled only if Two-way SSL is selected.
SSL Private Key The directory path where the .pem file containing the client's SSL private key is located. This field is enabled only if Two-way SSL is selected.
Client Private Key Password The password of the private key file that is specified in SSL Private Key. This field is enabled only if Two-way SSL is selected.


Hadoop Client

HDFS locations can only be accessed through HVR running on Linux or Windows, and it is not required to run HVR installed on the Hadoop Namenode although it is possible to do so. The Hadoop client should be present on the machine from which HVR will access the HDFS. For more information about installing Hadoop client, refer to Apache Hadoop Releases.

Hadoop Client Configuration

The following are required on the machine from which HVR connects to HDFS:

  • Hadoop 2.4.1 or newer client libraries with Java 7 Runtime Environment. For downloading Hadoop, refer to Apache Hadoop Releases.
  • Set the environment variable $JAVA_HOME to the Java installation directory.
  • Set the environment variable $HADOOP_COMMON_HOME or $HADOOP_HOME or $HADOOP_PREFIX to the Hadoop installation directory, or the hadoop command line client should be available in the path.
Note: Since the binary distribution available in Hadoop website lacks Windows-specific executables, a warning about unable to locate winutils.exe is displayed. This warning can be ignored for using Hadoop library for client operations to connect to a HDFS server using HVR. However, the performance on integrate location would be poor due to this warning, so it is recommended to use a Windows-specific Hadoop distribution to avoid this warning. For more information about this warning, refer to Hadoop Wiki and Hadoop issue HADOOP-10051.

Verifying Hadoop Client Installation

To verify the Hadoop client installation,

  1. The HADOOP_HOME/bin directory in Hadoop installation location should contain the hadoop executables in it.
  2. Execute the following commands to verify Hadoop client installation:
  3. $JAVA_HOME/bin/java -version
    $HADOOP_HOME/bin/hadoop version
    $HADOOP_HOME/bin/hadoop classpath
    
  4. If the Hadoop client installation is verified successfully then execute the following command to verify the connectivity between HVR and HDFS:
  5. $HADOOP_HOME/bin/hadoop fs -ls hdfs://<cluster>/
    

Client Configuration Files

Client configuration files are not required for HVR to perform replication, however, they can be useful for debugging. Client configuration files contain settings for different services like HDFS or HBASE. If the HVR integrate machine is not part of the cluster, it is recommended to download the configuration files for the cluster so that the Hadoop client knows how to connect to HDFS.

The client configuration files for Cloudera Manager or Ambari for Hortonworks can be downloaded from the respective cluster manager's web interface. For more information about downloading client configuration files, search for "Client Configuration Files" in the respective documentation for Cloudera and Hortonworks.

Hive External Table

HVR allows you to create Hive external tables above HDFS files which are only used during compare. The Hive ODBC connection can be enabled for Azure DLS in the location creation screen by selecting the Hive External Tables field (see section Location Connection).

ODBC Connection

HVR uses ODBC connection to the Hadoop cluster for which it requires the ODBC driver (Amazon ODBC 1.1.1 or HortonWorks ODBC 2.1.2 and above) for Hive installed on the machine (or in the same network).

The Amazon and HortonWorks ODBC drivers are similar and compatible to work with Hive 2.x release. However, it is recommended to use the Amazon ODBC driver for Amazon Hive and the Hortonworks ODBC driver for HortonWorks Hive.

By default, HVR uses Amazon ODBC driver for connecting to Hadoop. To use the Hortonworks ODBC driver the following action definition is required:
For Linux,

Group Table Action
Azure DLS * Environment /Name=HVR_ODBC_CONNECT_STRING_DRIVER /Value=Hortonworks Hive ODBC Driver 64-bit

For Windows,

Group Table Action
Azure DLS * Environment /Name=HVR_ODBC_CONNECT_STRING_DRIVER /Value=Hortonworks Hive ODBC Driver

Channel Configuration

For the file formats (CSV, JSON, and AVRO) the following action definitions are required to handle certain limitations of the Hive deserialization implementation during Bulk or Row-wise Compare:

  • For CSV,
  • Group Table Action
    Azure DLS * FileFormat /NullRepresentation=\\N
    Azure DLS * TableProperties /CharacterMapping="\x00>\\0;\n>\\n;\r>\\r;">\""
    Azure DLS * TableProperties /MapBinary=BASE64
  • For JSON,
  • Group Table Action
    Azure DLS * TableProperties /MapBinary=BASE64
    Azure DLS * FileFormat /JsonMode=ROW_FRAGMENTS
  • For Avro,
  • Group Table Action
    Azure DLS * FileFormat /AvroVersion=v1_8

    v1_8 is the default value for FileFormat /AvroVersion, so it is not mandatory to define this action.