Big Data Appliance X3-2 In-Rack Expansion
 
Big Data Appliance X5-2 Starter Rack
 
Big Data Appliance X5-2 Hardware
 
Big Data Appliance X4-2 Starter Rack
 
Big Data Appliance Integrated Software
 
Big Data Appliance X3-2 Full Rack
 
Big Data Appliance X5-2 In-Rack Expansion
 
Big Data Appliance X4-2 Full Rack
 
Big Data Appliance X3-2 Hardware
 
Big Data Appliance X4-2 Hardware
 
Big Data Appliance X4-2 In-Rack Expansion
 
Big Data Appliance X3-2 Starter Rack
 
Big Data Appliance X5-2 Full Rack
 

PLA-Support>Eng Systems>BDA>Big Data Appliance>DB: BDA_EST
 

In this Document

Purpose

Scope

Details

Overview

Single Rack Cluster

Multirack Clusters

In-Rack Expansions Software Layout

Oracle NoSQL Database Installation

References

Applies to:

Purpose

This information details the software layout for CDH5 services on Oracle Big Data Appliance (Sun Fire X4270 M2, Oracle Big Data Appliance X3-2, Oracle Big Data Appliance X4-2, and Oracle Big Data Appliance X5-2) Starter Rack, Full Rack, and Multirack installations when using Mammoth version V4.3.0/V4.4.0.

Scope

 This information is intended for the BDA Administrator as well as for developers.

Details

Overview

Individual services run only on designated nodes of a Hadoop cluster. Services that run on all nodes run on all racks of a multi-rack installation. The tables provided detail where CDH services run in a single rack cluster and multi-rack cluster. In addition the tables show the service layout for Starter Rack installations of 6 nodes. See the documentation.

Single Rack Cluster

The table below (table 1) shows the service type, the service roles, node name, and where the services run on the primary rack for the servers in the Big Data Cluster.

Table 1 - Service Locations for One or More CDH Clusters in a Single Rack

Note: If Oracle Big Data Discovery is installed, the NodeManager and DataNode on Node05 of the primary cluster are decomissioned.

Further information can be found in Where do NodeManagers Run in BDA V4.2 MultiRack CDH Clusters? (Doc ID 2019782.1).

Or the documentation for Multirack CDH Clusters.

Multirack Clusters

When multiple racks are configured as a single CDH cluster, some critical services are installed on the second rack. There can be variations in the distribution of services between the first and second racks in different multirack clusters. The two scenarios that account for these differences are:

• The cluster spanned multiple racks in its original configuration. 

          The resulting service locations across the first and second rack are described in Table 2 and Table 3. In this case, the JournalNode, Mysql Primary, ResourceManager and Zookeeper are installed on node 2 of the first rack.

• The cluster was single-rack in the original configuration and was extended later.

          The resulting service locations across the first and second rack are described in Table 4 and Table 5 . In this case, the JournalNode, Mysql Primary, ResourceManager, and Zookeeper are installed on node 3 of the first rack.

Note: If Oracle Big Data Discovery is installed, the NodeManager and DataNode on node 5 in the first rack of the primary cluster are decommissioned.

The second rack must have at least two nodes.

Critical services on the first server of the second rack:

•     Balancer
•     Failover Controller
•     Journal Node
•     NameNode
•     ZooKeeper Server

The DataNode, Cloudera Manager agent, and Puppet services also run on this server.

Table 2 - Service Locations in the First Rack (When the Cluster Started as Multirack Cluster)

* Nodenn includes the servers in additional racks.

Table 3 - Service Locations in the Second Rack (When the Cluster Started as Multirack Cluster)

* Nodenn includes the servers in additional racks.

Table 4 - First Rack Service Locations (When a Single-Rack Cluster is Extended)

Table 5 shows the service locations in the second rack of a cluster originally configured as a single-rack cluster and subsequently extended.

Table 5 - Second Rack Service Locations (When a Single-Rack Cluster is Extended)

Note: When expanding a cluster from one to two racks, Mammoth moves all critical services from nodes 2 and 4 of the first rack to nodes 1 and 2 of the second rack. Nodes 2 and 4 of the first rack become noncritical nodes.

In-Rack Expansions Software Layout

The flexibility of an In-Rack Expansion provides the ability to configure the cluster with different options. For In-Rack Expansions the following installation options are available:

1. With a Starter Rack and one In-Rack Expansion

a. Create a new cluster on the new six nodes. In this installation the new 6-node cluster will have the standard 6-node layout as shown in the Oracle Big Data Appliance Starter Rack Column above.

b. Extend the existing 6-node cluster on to the new servers to become a 12-node cluster. In this installation the new 12-node cluster will have the same layout as a Oracle Big Data Appliance Full Rack cluster. Note that in this option there will be no NodeManagers on nodes 1 and 2.

2. With a Starter Rack and two In-Rack Expansions the following options are available:

a. Create three 6-node clusters. In this case the software layout / service roles will be the same as in the Oracle Big Data Appliance Starter Rack column.

b. Create two 9-node clusters. In this case the software layout / service roles will be the same as in the Oracle Big Data Appliance Full Rack column.

c. Create a 6-node cluster and a 12-node cluster. In this case the software layout / service roles will be the same as in the Oracle Big Data Appliance Starter Rack column for the 6-node cluster. The software layout / service roles will be the same as in the Oracle Big Data Appliance Full Rack column for the 12-node cluster.

d. Create a full 18-node cluster. In this case the software layout / service roles will be the same as in the Oracle Big Data Appliance Full Rack column.

3. With a Starter Rack and an extension of 3-nodes:

It is possible to extend the cluster by only 3-nodes. Once this is done and the cluster becomes a 9-node cluster the 6-node Starter Rack configuration software layout / service roles will become the same as the Oracle Big Data Appliance Full Rack cluster. The NodeManagers on node01 and node02 will no longer be there. There will then be 7 NodeManagers instead of 6 NodeManagers. There will also be an increase in stability and suitability for a production cluster since it will no longer be running NodeManagers on the name nodes.

Oracle NoSQL Database Installation

As of Mammoth 2.2.0 installing a cluster with both Hadoop and NoSQL Database is no longer supported. Instead installing either a NoSQL Database cluster or a Hadoop cluster is supported on Oracle Big Data Appliance Hardware. A NoSQL Database cluster can be 3, 6, 9, 12, 15 or 18 servers in size. It cannot currently be extended once it is installed. Note that the configuration utility will not currently handle a 3 node NoSQL cluster directly.  Follow up in a SR with Oracle Support to achieve that configuration.

When upgrading an older version cluster that has NoSQL DB installed in addition to Hadoop, it will be required to remove the NoSQL DB installation first before doing an upgrade. To facilitate this use the mammoth-reconfig remove nosql option on an Oracle Big Data Appliance version 2.1 or earlier cluster to remove the current NoSQL DB installed alongside Hadoop. This option is available after extracting the Mammoth version 2.2.0 bundle but before running "mammoth -p" to do the upgrade.

References

This solution has no attachment