Operational Changes

Note the following functional changes to existing commands in Version 6.0.

MapR-DB

Mapr-DB TABLE API

MapR-DB Java TABLE API is deprecated and replaced by the DocumentStore OJAI API

MapR-DB JSON Client Connection

The client connection to the database has changed to the Connectionand Driver OJAI APIs

MapR-DB JSON Table Create

The insertionorder option for MapR-DB JSON tables is not supported. When upgrading from MapR 5.x, all existing MapR-DB JSON tables with insertionorder option set to true will show a new behavior where insertionorder is set to false. This impacts maprcli table create, maprcli table info, and dbshell create.

Library change

There was a library change for Hive. See Hive 2.1.1-1710 Release Notes

Secondary Indexes and SSDs

Solid-state drives (SSDs) are recommended when using MapR-DB JSON with secondary indexes. See Cluster Hardware.

MapR-FS

Chunk Size If chunk size is zero (0), when an application makes a request for block size, MapR will now return 2^62 (4611686018427387904); however, hadoop mfs commands will continue to display 0 for chunk size.

core-site.xml core-site.xml is now located in in /opt/mapr/hadoop/hadoop-<version>/etc/hadoop/ The previous location was /opt/mapr/hadoop/hadoop/conf

Default Database

Configure.sh -defaultdb = HBase (hbase) is no longer supported. The only supported option is MapR-DB (maprdb).

Mapr-ES (Streams)

The message offset in a partition starts from zero (0). If you are upgrading and do not enable the MapR-DB/ES feature, mfs.feature.db.streams.v6.support, the message offset in a partition starts from one (1).

Spyglass

MapR-ES (Streams) is now the default mechanism through which metrics flow from collectd to OpenTSDB.

MapR Control System

In prior releases, the mapr-webserver package contained both the MCS UI static files and the server running the Java application. Starting from v6.0, the UI static files are in mapr-webserver and the mapr-apiserver runs the server that handles the queries. The apiserver allows you to perform cluster administration programmatically. For more information, see Setting Up the MapR Control System

MapR Installer

Package dependencies have changed. Before using the installer, make sure that the packages described in MapR Installer Prerequisites and Guidelines are installed. These packages are downloaded for you when you use the MapR Installer, but must be installed manually before you install without using the MapR Installer.

MapR Installer Stanzas

MapR Installer 1.7 (which includes MapR Installer Stanzas) no longer supports the scaled_hosts: parameter for adding nodes to an on-premise cluster. Instead, you need to use the scaled_hosts2: parameter. See Extending a Cluster by Adding Nodes.

Deprecated Features

  • MRv1

    MapReduce version 1 is no longer supported. If you were previously using MapReduce version 1, verify the mode your cluster is running in and uninstall packages related to MapReduce version 1 before upgrading to MapR 6.0. For more information, see Migrating from MapReduce Version 1 to MapReduce Version 2. As part of this deprecation, the ExpressLane feature is no longer available

  • Ganglia

    Ganglia is no longer supported. However, you can view cluster metrics in MCS dashboards or custom MapR Monitoring dashboards. For more information, see Using MapR Monitoring (Spyglass Initiative).

  • Mahout
  • Storm
  • Sahara
  • Hadoop Commands

    The following hadoop commands are deprecated and replaced by mapred command. See Supported Commands for Hadoop 2.x

    • hadoop job
    • hadoop mradmin
    • hadoop pipes
    • hadoop queue
  • HBase
  • Node metrics
  • maprcli commands: job and task
    • The maprcli job changepriority, job kill, and job table commands are no longer supported
    • The maprcli task command including task failattempt, task killattempt, and task table, is no longer supported.
    • You may still see references to these commands in help text.

MapR Client

For the Windows client, you do not need to configure the core-site.xml file with the UID, GID, and user name of the cluster user to access a secure cluster because the username information will be available through the ticket. You must configure the core-site.xml file for accessing non-secure cluster only. See Configuring MapR Client User on Windows