Changes in Oracle Big Data Appliance Release 4.12

J.2 Changes in Oracle Big Data Appliance Release 4.12

Release 4.12 includes the following software revisions and new features.

Updated Software

Oracle Linux 7 with UEK4 for new clusters (as well as continued support for Oracle Linux 6 with UEK2 for cluster upgrades).
Cloudera Enterprise 5.14, including CDH 5.14.2, Cloudera Manager 5.14.3, and Key Trustee 5.14.0.
Oracle NoSQL Database Enterprise Edition 18.1.7
Oracle Big Data Connectors 4.12
MySQL Enterprise Edition 5.7.21
Java JDK 8u171
Oracle R Advanced Analytics for Hadoop (ORAAH) 2.8.0
Oracle's R Distribution (ORD) 3.3.0
Oracle NoSQL Community Edition 18.1.7

Other Software

Big Data SQL 3.2
Oracle Big Data Spatial & Graph 2.5
ODI Agent 12.2.1.3.0
Perfect Balance 2.10.0

Note:
Perfect Balance is deprecated in this release of Oracle Big Data Appliance and will be de-supported in a future release.

Kafka Clusters on Dedicated Oracle Big Data Appliance Nodes

Oracle Big Data Appliance now supports dedicated Kafka clusters using Cloudera’s CDK Powered By Apache Kafka 3.0.

Each cluster is managed by its own instance of Cloudera Manager. Separating Kafka from CDH in this way provides greater flexibility and easier management.

You can configure Kafka clusters though the Oracle Big Data Appliance Configuration Generation Utility.

Kafka clusters can be configured to use either Ethernet or InfiniBand. For security, both AD Kerberos and MIT Kerberos are options.

Note:

Cluster extension (./mammoth -e) is not available for Kafka clusters in this release.

Oracle Big Data Manager

Oracle Big Data Manager is a browser-based tool that gives you broad capabilities to manage data across your enterprise. You can use it to connect to and interconnect a range of supported Oracle and non-Oracle data storage providers, including Oracle Database, Oracle Object Store, MySQL, as well as Hadoop, S3, and GitHub. After you register storage providers with Big Data Manager, you can preview data and (depending upon the accessibility of each storage provider) compare, copy, and move data between them. With a Hadoop storage provider, you can also move data internally within HDFS, do data import/export and analytics with Apache Zeppelin, and import data into Hive tables. You can also upload data from your local computer to a selected storage provider.

At this time, Oracle Big Data Manager is not supported for clusters secured by Active Directory Kerberos or by MIT Kerberos when deployed with Key Distribution Center hosts that are external to Oracle Big Data Appliance.

Full Clusters on Oracle Linux 7

New racks are now delivered with Oracle Linux 7.

A new Oracle Linux 7 base image is available for re-imaging X3–2L to X6–2L servers.

Note:

There is no in-place procedure for upgrading servers from Oracle Linux 6 to Oracle Linux 7 at this time.
Sun Fire X4270 M2 servers cannot be re-imaged to Oracle Linux 7.

See Oracle Big Data Appliance Patch Set Master Note (Doc ID 1485745.1) in My Oracle Support (support.oracle.com) for details.

Oracle Big Data Appliance X7–2 SSDs Used to Store Journal Node Metadata and Zookeeper Data

This data is now stored on the two X7–2 SSDs rather than on disk. This provides better performance for highly loaded master nodes.

High Availability for Hue, Sentry, Hive Metastore, and AD Kerberos for Kafka Clusters

Oracle Big Data Appliance 4.12 now includes HA (High Availability) Hue, Sentry, and Hive Metastore.

In addition, the Hue Load Balancer service is added to each node that runs the Hue service. After the upgrade, users who are connected directly to the Hue service are prompted to connect to the Hue Load Balancer instead.

Note:

If the Mammoth upgrade to Oracle Big Data Appliance release 4.12 detects that non-HA Sentry is already present in the cluster, it displays a message to notify you that enabling HA for Sentry in requires a short shutdown of all Sentry-dependent services, regardless of whether you choose a rolling upgrade or a conventional upgrade. You can still select the rolling upgrade option when prompted, but the rolling upgrade will not be able to sustain full availability for Sentry and related services.

To enable Hive Metastore HA, the Hive Metastore Server Default Group is changed to org.apache.hadoop.hive.thrift.DBTokenStore.

Support for Big Data Discovery 1.6

Big Data Discovery 1.6 is now supported. You can download the installation files from the Oracle Software Delivery Cloud (also known as eDelivery).

The procedures for installing and upgrading Big Data Discovery on Oracle Big Data Appliance have not changed with release 1.6. See Installing Oracle Big Data Discovery in this guide for installation instructions. (This procedure differs from the one described in the Oracle Big Data Discovery Installation Guide.)

For upgrades, follow the instructions in the Oracle Big Data Discovery Upgrade Guide.

Other Changes

Three NTP Servers Recommended

Oracle strongly recommends (but does not require) that you add at least three NTP servers when configuring the network properties for Oracle Big Data Appliance. The reasons for this are:

If there is only one NTP server and it fails, other problems may occur with the cluster.
If there are two servers, they may not be synchronized.
Three servers increase the likelihood that at least two will be in agreement, and if so, this is the time that is used.