VMware vSphere Big Data Extensions

Command-Line Interface Guide

vSphere Big Data Extensions 2.0

This document supports the version of each product listed and supports all subsequent versions until the document is replaced by a new edition. To check for more recent editions of this document, see http://www.vmware.com/support/pubs.

EN-001513-00

VMware vSphere Big Data Extensions Command-Line Interface Guide

You can find the most up-to-date technical documentation on the VMware Web site at:

http://www.vmware.com/support/

The VMware Web site also provides the latest product updates.

If you have comments about this documentation, submit your feedback to:

docfeedback@vmware.com

(http://creativecommons.org/licenses/by-nd/3.0/us/legalcode).

VMware, Inc.

3401 Hillview Ave. Palo Alto, CA 94304 www.vmware.com

2 VMware, Inc.

About This Book

VMware vSphere Big Data Extensions Command-Line Interface Guide describes how to use the Serengeti Command-Line Interface (CLI) to manage the vSphere resources that you use to create Hadoop and HBase clusters, and how to create, manage, and monitor Hadoop and HBase clusters with the Serengeti CLI.

VMware vSphere Big Data Extensions Command-Line Interface Guide also describes how to perform Hadoop and HBase operations with the Serengeti CLI, and provides cluster specification and Serengeti CLI command references.

Intended Audience

This guide is for system administrators and developers who want to use Serengeti to deploy and manage Hadoop clusters. To successfully work with Serengeti, you should be familiar with Hadoop and VMware vSphere®.

VMware Technical Publications Glossary

VMware Technical Publications provides a glossary of terms that might be unfamiliar to you. For definitions of terms as they are used in VMware technical documentation, go to

http://www.vmware.com/support/pubs.

®

VMware, Inc.

5

VMware vSphere Big Data Extensions Command-Line Interface Guide

6 VMware, Inc.

Using the Serengeti Remote

Command-Line Interface Client 1

The Serengeti Remote Command-Line Interface Client lets you access the Serengeti Management Server to deploy, manage, and use Hadoop.

Access the Serengeti CLI By Using the Remote Command-Line Interface Client

You can access the Serengeti Command-Line Interface (CLI) to perform Serengeti administrative tasks with the Serengeti Remote CLI Client.

IMPORTANT You can only run Hadoop commands from the Serengeti CLI on a cluster running the Apache Hadoop 1.2.1 distribution. To use the command-line to run Hadoop administrative commands for clusters running other Hadoop distributions, such as cfg, fs, mr, pig, and hive, use a Hadoop client node to run these commands.

Prerequisites

Use the vSphere Web Client to log in to the vCenter Server on which you deployed the Serengeti vApp.

n

Verify that the Serengeti vApp deployment was successful and that the Management Server is running.

n

Verify that you have the correct password to log in to Serengeti CLI. See the VMware vSphere Big Data

n

Extensions Administrator's and User's Guide.

The Serengeti CLI uses its vCenter Server credentials.

Verify that the Java Runtime Environment (JRE) is installed in your environment and that its location is

n

in your PATH environment variable.

Procedure

1 Open a Web browser to connect to the Serengeti Management Server cli directory.

http://ip_address/cli

2 Download the ZIP file for your version and build.

The filename is in the format VMware-Serengeti-cli-version_number-build_number.ZIP.

3 Unzip the download.

The download includes the following components.

The serengeti-cli-version_number JAR file, which includes the Serengeti Remote CLI Client.

n

The samples directory, which includes sample cluster configurations.

n

Libraries in the lib directory.

n

VMware, Inc.

7

VMware vSphere Big Data Extensions Command-Line Interface Guide

4 Open a command shell, and change to the directory where you unzipped the package.

5 Change to the cli directory, and run the following command to enter the Serengeti CLI.

For any language other than French of German, run the following command.

n

java -jar serengeti-cli-version_number.jar

For French or German languages, which use code page 850 (CP 850) language encoding when

n

running the Serengeti CLI from a Windows command console, run the following command.

java -Dfile.encoding=cp850 -jar serengeti-cli-version_number.jar

6 Connect to the Serengeti service.

You must run the connect host command every time you begin a CLI session, and again after the 30 minute session timeout. If you do not run this command, you cannot run any other commands.

a Run the connect command.

connect --host xx.xx.xx.xx:8443

b At the prompt, type your user name, which might be different from your login credentials for the

Serengeti Management Server.

NOTE If you do not create a user name and password for the Serengeti Command-Line Interface Client, you can use the default vCenter Server administrator credentials. The Serengeti CommandLine Interface Client uses the vCenter Server login credentials with read permissions on the Serengeti Management Server.

c At the prompt, type your password.

A command shell opens, and the Serengeti CLI prompt appears. You can use the help command to get help with Serengeti commands and command syntax.

To display a list of available commands, type help.

n

To get help for a specific command, append the name of the command to the help command.

n

help cluster create

Press Tab to complete a command.

n

8 VMware, Inc.

Managing vSphere Resources for

Hadoop and HBase Clusters 2

Big Data Extensions lets you manage the resource pools, datastores, and networks that you use in the Hadoop and HBase clusters that you create.

Add a Resource Pool with the Serengeti Command-Line Interface on page 10

n

You add resource pools to make them available for use by Hadoop clusters. Resource pools must be located at the top level of a cluster. Nested resource pools are not supported.

Remove a Resource Pool with the Serengeti Command-Line Interface on page 10

n

You can remove resource pools from Serengeti that are not in use by a Hadoop cluster. You remove resource pools when you do not need them or if you want the Hadoop clusters you create in the Serengeti Management Server to be deployed under a different resource pool. Removing a resource pool removes its reference in vSphere. The resource pool is not deleted.

Add a Datastore with the Serengeti Command-Line Interface on page 10

n

You can add shared and local datastores to the Serengeti server to make them available to Hadoop clusters.

Remove a Datastore with the Serengeti Command-Line Interface on page 11

n

You can remove any datastore from Serengeti that is not referenced by any Hadoop clusters. Removing a datastore removes only the reference to the vCenter Server datastore. The datastore itself is not deleted.

VMware, Inc.

Add a Network with the Serengeti Command-Line Interface on page 11

n

You add networks to Serengeti to make their IP addresses available to Hadoop clusters. A network is a port group, as well as a means of accessing the port group through an IP address.

Reconfigure a Static IP Network with the Serengeti Command-Line Interface on page 12

n

You can reconfigure a Serengeti static IP network by adding IP address segments to it. You might need to add IP address segments so that there is enough capacity for a cluster that you want to create.

Remove a Network with the Serengeti Command-Line Interface on page 12

n

You can remove networks from Serengeti that are not referenced by any Hadoop clusters. Removing an unused network frees the IP addresses for reuse.

9

VMware vSphere Big Data Extensions Command-Line Interface Guide

Add a Resource Pool with the Serengeti Command-Line Interface

You add resource pools to make them available for use by Hadoop clusters. Resource pools must be located at the top level of a cluster. Nested resource pools are not supported.

When you add a resource pool to Big Data Extensions it symbolically represents the actual vSphere resource pool as recognized by vCenter Server. This symbolic representation lets you use the Big Data Extensions resource pool name, instead of the full path of the resource pool in vCenter Server, in cluster specification files.

Prerequisites

Deploy Big Data Extensions.

Procedure

1 Access the Serengeti Command-Line Interface client.

2 Run the resourcepool add command.

The --vcrp parameter is optional.

This example adds a Serengeti resource pool named myRP to the vSphere rp1 resource pool that is contained by the cluster1 vSphere cluster.

resourcepool add --name myRP --vccluster cluster1 --vcrp rp1

What to do next

After you add a resource pool to Big Data Extensions, do not rename the resource pool in vSphere. If you rename it, you cannot perform Serengeti operations on clusters that use that resource pool.

Remove a Resource Pool with the Serengeti Command-Line Interface

You can remove resource pools from Serengeti that are not in use by a Hadoop cluster. You remove resource pools when you do not need them or if you want the Hadoop clusters you create in the Serengeti Management Server to be deployed under a different resource pool. Removing a resource pool removes its reference in vSphere. The resource pool is not deleted.

Procedure

1 Access the Serengeti Command-Line Interface client.

2 Run the resourcepool delete command.

If the command fails because the resource pool is referenced by a Hadoop cluster, you can use the

resourcepool list command to see which cluster is referencing the resource pool.

This example deletes the resource pool named myRP.

resourcepool delete --name myRP

Add a Datastore with the Serengeti Command-Line Interface

You can add shared and local datastores to the Serengeti server to make them available to Hadoop clusters.

Procedure

1 Access the Serengeti CLI.

10 VMware, Inc.

Chapter 2 Managing vSphere Resources for Hadoop and HBase Clusters

2 Run the datastore add command.

This example adds a new, local storage datastore named myLocalDS. The --spec parameter’s value,

local*, is a wildcard specifying a set of vSphere datastores. All vSphere datastores whose names begin

with “local” are added and managed as a whole by Serengeti.

datastore add --name myLocalDS --spec local* --type LOCAL

What to do next

After you add a datastore to Big Data Extensions, do not rename the datastore in vSphere. If you rename it, you cannot perform Serengeti operations on clusters that use that datastore.

Remove a Datastore with the Serengeti Command-Line Interface

You can remove any datastore from Serengeti that is not referenced by any Hadoop clusters. Removing a datastore removes only the reference to the vCenter Server datastore. The datastore itself is not deleted.

You remove datastores if you do not need them or if you want to deploy the Hadoop clusters that you create in the Serengeti Management Server under a different datastore.

Procedure

1 Access the Serengeti CLI.

2 Run the datastore delete command.

If the command fails because the datastore is referenced by a Hadoop cluster, you can use the datastore

list command to see which cluster is referencing the datastore.

This example deletes the myDS datastore.

datastore delete --name myDS

Add a Network with the Serengeti Command-Line Interface

You add networks to Serengeti to make their IP addresses available to Hadoop clusters. A network is a port group, as well as a means of accessing the port group through an IP address.

Prerequisites

If your network uses static IP addresses, be sure that the addresses are not occupied before you add the network.

Procedure

1 Access the Serengeti CLI.

VMware, Inc. 11

VMware vSphere Big Data Extensions Command-Line Interface Guide

2 Run the network add command.

This example adds a network named myNW to the 10PG vSphere port group. Virtual machines that use this network use DHCP to obtain the IP addresses.

network add --name myNW --portGroup 10PG --dhcp

This example adds a network named myNW to the 10PG vSphere port group. Hadoop nodes use addresses in the 192.168.1.2-100 IP address range, the DNS server IP address is 10.111.90.2, the gateway address is 192.168.1.1, and the subnet mask is 255.255.255.0.

network add --name myNW --portGroup 10PG --ip 192.168.1.2-100 --dns 10.111.90.2 --gateway

192.168.1.1 --mask 255.255.255.0

To specify multiple IP address segments, use multiple strings to express the IP address range in the format xx.xx.xx.xx-xx[,xx]*. For example:

xx.xx.xx.xx-xx, xx.xx.xx.xx-xx, single_ip, single_ip

What to do next

After you add a network to Big Data Extensions, do not rename it in vSphere. If you rename the network, you cannot perform Serengeti operations on clusters that use that network.

Reconfigure a Static IP Network with the Serengeti Command-Line Interface

You can reconfigure a Serengeti static IP network by adding IP address segments to it. You might need to add IP address segments so that there is enough capacity for a cluster that you want to create.

If the IP range that you specify includes IP addresses that are already in the network, Serengeti ignores the duplicated addresses. The remaining addresses in the specified range are added to the network. If the network is already used by a cluster, the cluster can use the new IP addresses after you add them to the network. If only part of the IP range is used by a cluster, the unused IP address can be used when you create a new cluster.

Prerequisites

If your network uses static IP addresses, be sure that the addresses are not occupied before you add the network.

Procedure

1 Access the Serengeti CLI.

2 Run the network modify command.

This example adds IP addresses from 192.168.1.2 to 192.168.1.100 to a network named myNetwork.

network modify --name myNetwork --addIP 192.168.1.2-100

Remove a Network with the Serengeti Command-Line Interface

You can remove networks from Serengeti that are not referenced by any Hadoop clusters. Removing an unused network frees the IP addresses for reuse.

Procedure

1 Access the Serengeti CLI.

12 VMware, Inc.

Chapter 2 Managing vSphere Resources for Hadoop and HBase Clusters

2 Run the network delete command.

network delete --name network_name

If the command fails because the network is referenced by a Hadoop cluster, you can use the network

list --detail command to see which cluster is referencing the network.

VMware, Inc. 13

VMware vSphere Big Data Extensions Command-Line Interface Guide

14 VMware, Inc.

Creating Hadoop and HBase Clusters 3

Big Data Extensions lets you create and deploy Hadoop and HBase clusters. A Hadoop or HBase cluster is a special type of computational cluster designed specifically for storing and analyzing large amounts of unstructured data in a distributed computing environment.

The resource requirements are different for clusters created with the Serengeti Command-Line Interface and the Big Data Extensions plug-in for the vSphere Web Client because the clusters use different default templates. The default clusters created through the Serengeti Command-Line Interface are targeted for Project Serengeti users and proof-of-concept applications, and are smaller than the Big Data Extensions plug-in templates, which are targeted for larger deployments for commercial use.

Additionally, some deployment configurations require more resources than other configurations. For example, if you create a Greenplum HD 1.2 cluster, you cannot use the SMALL size virtual machine. If you create a default MapR or Greenplum HD cluster through the Serengeti Command-Line Interface, at least 550GB of storage and 55GB of memory are recommended. For other Hadoop distributions, at least 350GB of storage and 35GB of memory are recommended.

CAUTION When you create a cluster with Big Data Extensions, Big Data Extensions disables the cluster's virtual machine automatic migration. Although this prevents vSphere from automatically migrating the virtual machines, it does not prevent you from inadvertently migrating cluster nodes to other hosts by using the vCenter Server user interface. Do not use the vCenter Server user interface to migrate clusters. Performing such management functions outside of the Big Data Extensions environment can make it impossible for you to perform some Big Data Extensions operations, such as disk failure recovery.

VMware, Inc.

About Hadoop and HBase Cluster Deployment Types on page 17

n

Big Data Extensions lets you deploy several types of Hadoop and HBase clusters. You need to know about the types of clusters that you can create.

Serengeti’s Default Hadoop Cluster Configuration on page 18

n

For basic Hadoop deployments, such as proof of concept projects, you can use Serengeti’s default Hadoop cluster configuration for clusters that are created with the Command-Line Interface.

Create a Default Serengeti Hadoop Cluster with the Serengeti Command-Line Interface on page 18

n

You can create as many clusters as you want in your Serengeti environment, but your environment must meet all prerequisites.

Create a Cluster with a Custom Administrator Password with the Serengeti Command-Line Interface

n

on page 19

When you create a cluster, you can assign a custom administrator password to all the nodes in the cluster. Custom administrator passwords let you directly log in to the cluster's nodes instead of having to first log in to the Serengeti Management server.

15

VMware vSphere Big Data Extensions Command-Line Interface Guide

Create a Cluster with an Available Distribution with the Serengeti Command-Line Interface on

n

page 19

You can choose which Hadoop distribution to use when you deploy a cluster. If you do not specify a Hadoop distribution, the resulting cluster includes the default distribution, Apache Hadoop.

Create a Hadoop Cluster with Assigned Resources with the Serengeti Command-Line Interface on

n

page 20

By default, when you use Serengeti to deploy a Hadoop cluster, the cluster might contain any or all available resources: vCenter Server resource pool for the virtual machine's CPU and memory, datastores for the virtual machine's storage, and a network. You can assign which resources the cluster uses by specifying specific resource pools, datastores, and/or a network when you create the Hadoop cluster.

Create a Cluster with Multiple Networks with the Serengeti Command-Line Interface on page 21

n

When you create a cluster, you can distribute the management, HDFS, and MapReduce traffic to separate networks. You might want to use separate networks to improve performance or to isolate traffic for security reasons.

Create a MapReduce v2 (YARN) Cluster with the Serengeti Command-Line Interface on page 21

n

You can create MapReduce v2 (YARN) cluster with the Serengeti Command-Line Interface.

Create a Customized Hadoop or HBase Cluster with the Serengeti Command-Line Interface on

n

page 22

You can create clusters that are customized for your requirements, including the number of nodes, virtual machine RAM and disk size, the number of CPUs, and so on.

Create a Hadoop Cluster with Any Number of Master, Worker, and Client Nodes on page 23

n

You can create a Hadoop cluster with any number of master, worker, and client nodes.

Create a Data-Compute Separated Cluster with No Node Placement Constraints on page 24

n

You can create a cluster with separate data and compute nodes, without node placement constraints.

Create a Data-Compute Separated Cluster with Placement Policy Constraints on page 25

n

You can create a cluster with separate data and compute nodes, and define placement policy constraints to distribute the nodes among the virtual machines as you want.

Create a Compute-Only Cluster with the Serengeti Command-Line Interface on page 27

n

You can create compute-only clusters to run MapReduce jobs on existing HDFS clusters, including storage solutions that serve as an external HDFS.

Create a Basic Cluster with the Serengeti Command-Line Interface on page 29

n

You can create a basic cluster in your Serengeti environment. A basic cluster is a group of virtual machines provisioned and managed by Serengeti. Serengeti helps you to plan and provision the virtual machines to your specifications. You can use the basic cluster's virtual machines to install Big Data applications.

About Cluster Topology on page 31

n

You can improve workload balance across your cluster nodes, and improve performance and throughput, by specifying how Hadoop virtual machines are placed using topology awareness. For example, you can have separate data and compute nodes, and improve performance and throughput by placing the nodes on the same set of physical hosts.

Create a Cluster with Topology Awareness with the Serengeti Command-Line Interface on page 33

n

To achieve a balanced workload or to improve performance and throughput, you can control how Hadoop virtual machines are placed by adding topology awareness to the Hadoop clusters. For example, you can have separate data and compute nodes, and improve performance and throughput by placing the nodes on the same set of physical hosts.

16 VMware, Inc.

Chapter 3 Creating Hadoop and HBase Clusters

Create a Data-Compute Separated Cluster with Topology Awareness and Placement Constraints on

n

page 34

You can create clusters with separate data and compute nodes, and define topology and placement policy constraints to distribute the nodes among the physical racks and the virtual machines.

Serengeti’s Default HBase Cluster Configuration on page 36

n

HBase clusters are required for you to build big table applications. To run HBase MapReduce jobs, configure the HBase cluster to include JobTracker nodes or TaskTracker nodes.

Create a Default HBase Cluster with the Serengeti Command-Line Interface on page 36

n

Serengeti supports deploying HBase clusters on HDFS.

Create an HBase Cluster with vSphere HA Protection with the Serengeti Command-Line Interface on

n

page 37

You can create HBase clusters with separated Hadoop NameNode and HBase Master roles, and configure vSphere HA protection for the Masters.

About Hadoop and HBase Cluster Deployment Types

Big Data Extensions lets you deploy several types of Hadoop and HBase clusters. You need to know about the types of clusters that you can create.

You can create the following types of clusters.

Basic Hadoop Cluster

HBase Cluster

Data-Compute Separated Hadoop Cluster

Compute-Only Hadoop Cluster

Customized Cluster

You can create a simple Hadoop deployment for proof of concept projects and other small scale data processing tasks using the basic Hadoop cluster.

You can create an HBase cluster. To run HBase MapReduce jobs, configure the HBase cluster to include JobTracker or TaskTracker nodes.

You can separate the data and compute nodes in a Hadoop cluster, and you can control how nodes are placed on your environment's vSphere ESXi hosts.

You can create a compute-only cluster to run MapReduce jobs. Computeonly clusters run only MapReduce services that read data from external HDFS clusters and that do not need to store data.

You can use an existing cluster specification file to create clusters using the same configuration as your previously created clusters. You can also edit the file to customize the cluster configuration.

Hadoop Distributions Supporting MapReduce v1 and MapReduce v2 (YARN)

If the Hadoop distribution you use supports both MapReduce v1 and MapReduce v2 (YARN), the default Hadoop cluster configuration creates a MapReduce v2 cluster.

In addition, if you are using two different versions of a vendor's Hadoop distribution, and both versions support MapReduce v1 and MapReduce v2, the cluster you create using the latest version with the default Hadoop cluster will use MapReduce v2. Clusters you create with the earlier Hadoop version will use MapReduce v1. For example, if you have both Cloudera CDH 5 and CDH 4 installed within Big Data Extensions, clusters you create with CDH 5 will use MapReduce v2, and clusters you create with CDH 4 will use MapReduce v1.

VMware, Inc. 17

VMware vSphere Big Data Extensions Command-Line Interface Guide

Serengeti’s Default Hadoop Cluster Configuration

For basic Hadoop deployments, such as proof of concept projects, you can use Serengeti’s default Hadoop cluster configuration for clusters that are created with the Command-Line Interface.

The resulting cluster deployment consists of the following nodes and virtual machines:

One master node virtual machine with NameNode and JobTracker services.

n

Three worker node virtual machines, each with DataNode and TaskTracker services.

n

One client node virtual machine containing the Hadoop client environment: the Hadoop client shell,

n

Pig, and Hive.

Hadoop Distributions Supporting MapReduce v1 and MapReduce v2 (YARN)

If the Hadoop distribution you use supports both MapReduce v1 and MapReduce v2 (YARN), the default Hadoop cluster configuration creates a MapReduce v2 cluster.

In addition, if you are using two different versions of a vendor's Hadoop distribution, and both versions support MapReduce v1 and MapReduce v2, the cluster you create using the latest version with the default Hadoop cluster will use MapReduce v2. Clusters you create with the earlier Hadoop version will use MapReduce v1. For example, if you have both Cloudera CDH 5 and CDH 4 installed within Big Data Extensions, clusters you create with CDH 5 will use MapReduce v2, and clusters you create with CDH 4 will use MapReduce v1.

Create a Default Serengeti Hadoop Cluster with the Serengeti Command-Line Interface

You can create as many clusters as you want in your Serengeti environment, but your environment must meet all prerequisites.

Prerequisites

Deploy the Serengeti vApp.

n

Ensure that you have adequate resources allocated to run the Hadoop cluster.

n

To use any Hadoop distribution other than the provided Apache Hadoop, add one or more Hadoop

n

distributions. See the VMware vSphere Big Data Extensions Administrator's and User's Guide.

Procedure

1 Access the Serengeti CLI.

2 Deploy a default Serengeti Hadoop cluster on vSphere.

cluster create --name cluster_name

The only valid characters for cluster names are alphanumeric and underscores. When you choose the cluster name, also consider the applicable vApp name. Together, the vApp and cluster names must be < 80 characters.

During the deployment process, real-time progress updates appear on the command-line.

What to do next

After the deployment finishes, you can run Hadoop commands and view the IP addresses of the Hadoop node virtual machines from the Serengeti CLI.

18 VMware, Inc.

Chapter 3 Creating Hadoop and HBase Clusters

Create a Cluster with a Custom Administrator Password with the Serengeti Command-Line Interface

When you create a cluster, you can assign a custom administrator password to all the nodes in the cluster. Custom administrator passwords let you directly log in to the cluster's nodes instead of having to first log in to the Serengeti Management server.

Prerequisites

Deploy the Serengeti vApp.

n

Ensure that you have adequate resources allocated to run the Hadoop cluster.

n

To use any Hadoop distribution other than the provided Apache Hadoop, add one or more Hadoop

n

distributions. See the VMware vSphere Big Data Extensions Administrator's and User's Guide.

Procedure

1 Access the Serengeti CLI.

2 Run the cluster create command and include the --password parameter.

cluster create --name cluster_name --password

3 Enter your custom password, and enter it again.

Passwords are from 8 to 128 characters, and include only alphanumeric characters ([0-9, a-z, A-Z]) and the following special characters: _ @ # $ % ^ & *

Your custom password is assigned to all the nodes in the cluster.

Create a Cluster with an Available Distribution with the Serengeti Command-Line Interface

You can choose which Hadoop distribution to use when you deploy a cluster. If you do not specify a Hadoop distribution, the resulting cluster includes the default distribution, Apache Hadoop.

Prerequisites

Deploy the Serengeti vApp.

n

Ensure that you have adequate resources allocated to run the Hadoop cluster.

n

To use any Hadoop distribution other than the provided Apache Hadoop, add one or more Hadoop

n

distributions. See the VMware vSphere Big Data Extensions Administrator's and User's Guide.

Procedure

1 Access the Serengeti CLI.

VMware, Inc. 19

VMware vSphere Big Data Extensions Command-Line Interface Guide

2 Run the cluster create command, and include the --distro parameter.

The --distro parameter’s value must match a distribution name displayed by the distro list command.

NOTE To create an Apache Bigtop, Cloudera CDH4 and CDH5, or Pivotal PHD 1.1 cluster, you must configure a valid DNS and FQDN for the cluster's HDFS and MapReduce traffic. Without valid DNS and FQDN settings, the cluster creation process might fail or the cluster is created but does not function.

This example deploys a cluster with the Cloudera CDH distribution:

cluster create --name clusterName --distro cdh

This example creates a customized cluster named mycdh that uses the CDH4 Hadoop distribution, and is configured according to the /opt/serengeti/samples/default_cdh4_ha_and_federation_hadoop_cluster.json sample cluster specification file. In this sample file, nameservice0 and nameservice1 are federated. That is,

nameservice0 and nameservice1 are independent and do not require coordination with each other. The

NameNode nodes in the nameservice0 node group are HDFS2 HA enabled. In Serengeti, name node group names are used as service names for HDFS2.

cluster create --name mycdh --distro cdh4 -specFile /opt/serengeti/samples/default_cdh4_ha_hadoop_cluster.json

Create a Hadoop Cluster with Assigned Resources with the Serengeti Command-Line Interface

By default, when you use Serengeti to deploy a Hadoop cluster, the cluster might contain any or all available resources: vCenter Server resource pool for the virtual machine's CPU and memory, datastores for the virtual machine's storage, and a network. You can assign which resources the cluster uses by specifying specific resource pools, datastores, and/or a network when you create the Hadoop cluster.

Prerequisites

Deploy the Serengeti vApp.

n

Ensure that you have adequate resources allocated to run the Hadoop cluster.

n

To use any Hadoop distribution other than the provided Apache Hadoop, add one or more Hadoop

n

distributions. See the VMware vSphere Big Data Extensions Administrator's and User's Guide.

Procedure

1 Access the Serengeti CLI.

2 Run the cluster create command, and specify any or all of the command’s resource parameters.

This example deploys a cluster named myHadoop on the myDS datastore, under the myRP resource pool, and uses the myNW network for virtual machine communications.

cluster create --name myHadoop --rpNames myRP --dsNames myDS --networkName myNW

20 VMware, Inc.

Chapter 3 Creating Hadoop and HBase Clusters

Create a Cluster with Multiple Networks with the Serengeti CommandLine Interface

When you create a cluster, you can distribute the management, HDFS, and MapReduce traffic to separate networks. You might want to use separate networks to improve performance or to isolate traffic for security reasons.

For optimal performance, use the same network for HDFS and MapReduce traffic in Hadoop and Hadoop +HBase clusters. HBase clusters use the HDFS network for traffic related to the HBase Master and HBase RegionServer services.

IMPORTANT You cannot configure multiple networks for clusters that use the MapR Hadoop distribution.

Prerequisites

Deploy the Serengeti vApp.

n

Ensure that you have adequate resources allocated to run the Hadoop cluster.

n

To use any Hadoop distribution other than the provided Apache Hadoop, add one or more Hadoop

n

distributions. See the VMware vSphere Big Data Extensions Administrator's and User's Guide.

Procedure

1 Access the Serengeti CLI.

2 Run the cluster create command and include the --networkName, --hdfsNetworkName, and --

mapredNetworkName parameters.

cluster create --name cluster_name --networkName management_network [--hdfsNetworkName hdfs_network] [--mapredNetworkName mapred_network]

If you omit an optional network parameter, the traffic associated with that network parameter is routed on the management network that you specify by the --networkName parameter.

NOTE To create an Apache Bigtop, Cloudera CDH4 and CDH5, or Pivotal PHD 1.1 cluster, you must configure a valid DNS and FQDN for the cluster's HDFS and MapReduce traffic. Without valid DNS and FQDN settings, the cluster creation process might fail or the cluster is created but does not function.

The cluster's management, HDFS, and MapReduce traffic is distributed among the specified networks.

Create a MapReduce v2 (YARN) Cluster with the Serengeti CommandLine Interface

You can create MapReduce v2 (YARN) cluster with the Serengeti Command-Line Interface.

When you create a Hadoop cluster with the Serengeti Command-Line Interface, by default you create a MapReduce v1 cluster. To create a MapReduce v2 (YARN) cluster, create a cluster specification file modeled after the /opt/serengeti/samples/default_hadoop_yarn_cluster.json file, and specify the --specFile parameter and your cluster specification file in the cluster create ... command.

Prerequisites

Deploy the Serengeti vApp.

n

Ensure that you have adequate resources allocated to run the Hadoop cluster.

n

To use any Hadoop distribution other than the provided Apache Hadoop, add one or more Hadoop

n

distributions. See the VMware vSphere Big Data Extensions Administrator's and User's Guide.

VMware, Inc. 21

VMware vSphere Big Data Extensions Command-Line Interface Guide

Procedure

1 Access the Serengeti CLI.

2 Run the cluster create ... command.

This example creates a customized MapR v2 cluster according to the sample cluster specification file,

default_hadoop_yarn_cluster.json.

cluster create --name cluster_name --distro cdh4 -specFile /opt/serengeti/samples/default_hadoop_yarn_cluster.json

Create a Customized Hadoop or HBase Cluster with the Serengeti Command-Line Interface

You can create clusters that are customized for your requirements, including the number of nodes, virtual machine RAM and disk size, the number of CPUs, and so on.

The Serengeti package includes several annotated sample cluster specification files that you can use as models when you create your custom specification files.

In the Serengeti Management Server, the sample cluster specification files are

n

in /opt/serengeti/samples.

If you use the Serengeti Remote CLI client, the sample specification files are in the client directory.

n

Changing a node group role might cause the cluster creation process to fail. For example, workable clusters require a NameNode, so if there are no NameNode nodes after you change node group roles, you cannot create a cluster.

Prerequisites

Deploy the Serengeti vApp.

n

Ensure that you have adequate resources allocated to run the Hadoop cluster.

n

To use any Hadoop distribution other than the provided Apache Hadoop, add one or more Hadoop

n

distributions. See the VMware vSphere Big Data Extensions Administrator's and User's Guide.

Procedure

1 Create a cluster specification file to define the cluster's characteristics such as the node groups.

2 Access the Serengeti CLI.

3 Run the cluster create command, and specify the cluster specification file.

Use the full path to specify the file.

cluster create --name cluster_name --specFile full_path/spec_filename

NOTE To create an Apache Bigtop, Cloudera CDH4 and CDH5, or Pivotal PHD 1.1 cluster, you must configure a valid DNS and FQDN for the cluster's HDFS and MapReduce traffic. Without valid DNS and FQDN settings, the cluster creation process might fail or the cluster is created but does not function.

22 VMware, Inc.

Chapter 3 Creating Hadoop and HBase Clusters

Create a Hadoop Cluster with Any Number of Master, Worker, and Client Nodes

You can create a Hadoop cluster with any number of master, worker, and client nodes.

Prerequisites

Deploy the Serengeti vApp.

n

Ensure that you have adequate resources allocated to run the Hadoop cluster.

n

To use any Hadoop distribution other than the provided Apache Hadoop, add one or more Hadoop

n

distributions. See the VMware vSphere Big Data Extensions Administrator's and User's Guide.

Procedure

1 Create a cluster specification file to define the cluster's characteristics, including the node groups.

NOTE To create an Apache Bigtop, Cloudera CDH4 and CDH5, or Pivotal PHD 1.1 cluster, you must configure a valid DNS and FQDN for the cluster's HDFS and MapReduce traffic. Without valid DNS and FQDN settings, the cluster creation process might fail or the cluster is created but does not function.

In this example, the cluster has one master MEDIUM size virtual machine, five worker SMALL size virtual machines, and one client SMALL size virtual machine. The instanceNum attribute configures the number of virtual machines in a node.

{ "nodeGroups" : [ { "name": "master", "roles": [ "hadoop_namenode", "hadoop_jobtracker" ], "instanceNum": 1, "instanceType": "MEDIUM" }, { "name": "worker", "roles": [ "hadoop_datanode", "hadoop_tasktracker" ], "instanceNum": 5, "instanceType": "SMALL" }, { "name": "client", "roles": [ "hadoop_client", "hive", "hive_server", "pig" ], "instanceNum": 1,

VMware, Inc. 23

VMware vSphere Big Data Extensions Command-Line Interface Guide

"instanceType": "SMALL" } ] }

2 Access the Serengeti CLI.

3 Run the cluster create command, and specify the cluster specification file.

cluster create --name cluster_name --specFile full_path/spec_filename

Create a Data-Compute Separated Cluster with No Node Placement Constraints

You can create a cluster with separate data and compute nodes, without node placement constraints.

Prerequisites

Deploy the Serengeti vApp.

n

Ensure that you have adequate resources allocated to run the Hadoop cluster.

n

To use any Hadoop distribution other than the provided Apache Hadoop, add one or more Hadoop

n

distributions. See the VMware vSphere Big Data Extensions Administrator's and User's Guide.

Procedure

1 Create a cluster specification file to define the cluster's characteristics.

NOTE To create an Apache Bigtop, Cloudera CDH4 and CDH5, or Pivotal PHD 1.1 cluster, you must configure a valid DNS and FQDN for the cluster's HDFS and MapReduce traffic. Without valid DNS and FQDN settings, the cluster creation process might fail or the cluster is created but does not function.

In this example, the cluster has separate data and compute nodes, without node placement constraints. Four data nodes and eight compute nodes are created and put into individual virtual machines. The number of nodes is configured by the instanceNum attribute.

{ "nodeGroups":[ { "name": "master", "roles": [ "hadoop_namenode", "hadoop_jobtracker" ], "instanceNum": 1, "cpuNum": 2, "memCapacityMB": 7500, }, { "name": "data", "roles": [ "hadoop_datanode" ], "instanceNum": 4, "cpuNum": 1, "memCapacityMB": 3748, "storage": {

24 VMware, Inc.

"type": "LOCAL", "sizeGB": 50 } }, { "name": "compute", "roles": [ "hadoop_tasktracker" ], "instanceNum": 8, "cpuNum": 2, "memCapacityMB": 7500, "storage": { "type": "LOCAL", "sizeGB": 20 } }, { "name": "client", "roles": [ "hadoop_client", "hive", "pig" ], "instanceNum": 1, "cpuNum": 1, "storage": { "type": "LOCAL", "sizeGB": 50 } } ], "configuration": { } }

Chapter 3 Creating Hadoop and HBase Clusters

2 Access the Serengeti CLI.

3 Run the cluster create command and specify the cluster specification file.

cluster create --name cluster_name --specFile full_path/spec_filename

Create a Data-Compute Separated Cluster with Placement Policy Constraints

You can create a cluster with separate data and compute nodes, and define placement policy constraints to distribute the nodes among the virtual machines as you want.

CAUTION When you create a cluster with Big Data Extensions, Big Data Extensions disables the cluster's virtual machine automatic migration. Although this prevents vSphere from migrating the virtual machines, it does not prevent you from inadvertently migrating cluster nodes to other hosts by using the vCenter Server user interface. Do not use the vCenter Server user interface to migrate clusters. Performing such management functions outside of the Big Data Extensions environment might break the cluster's placement policy, such as the number of instances per host and the group associations. Even if you do not specify a placement policy, using vCenter Server to migrate clusters can break the default ROUNDROBIN placement policy constraints.

VMware, Inc. 25

VMware vSphere Big Data Extensions Command-Line Interface Guide

Prerequisites

Deploy the Serengeti vApp.

n

Ensure that you have adequate resources allocated to run the Hadoop cluster.

n

To use any Hadoop distribution other than the provided Apache Hadoop, add one or more Hadoop

n

distributions. See the VMware vSphere Big Data Extensions Administrator's and User's Guide.

Procedure

1 Create a cluster specification file to define the cluster's characteristics, including the node groups and

placement policy constraints.

NOTE To create an Apache Bigtop, Cloudera CDH4 and CDH5, or Pivotal PHD 1.1 cluster, you must configure a valid DNS and FQDN for the cluster's HDFS and MapReduce traffic. Without valid DNS and FQDN settings, the cluster creation process might fail or the cluster is created but does not function.

In this example, the cluster has data-compute separated nodes, and each node group has a

placementPolicy constraint. After a successful provisioning, four data nodes and eight compute nodes

are created and put into individual virtual machines. With the instancePerHost=1 constraint, the four data nodes are placed on four ESXi hosts. The eight compute nodes are put onto four ESXi hosts: two nodes on each ESXi host.

This cluster specification requires that you configure datastores and resource pools for at least four hosts, and that there is sufficient disk space for Serengeti to perform the necessary placements during deployment.

{ "nodeGroups":[ { "name": "master", "roles": [ "hadoop_namenode", "hadoop_jobtracker" ], "instanceNum": 1, "cpuNum": 2, "memCapacityMB": 7500, }, { "name": "data", "roles": [ "hadoop_datanode" ], "instanceNum": 4, "cpuNum": 1, "memCapacityMB": 3748, "storage": { "type": "LOCAL", "sizeGB": 50 }, "placementPolicies": { "instancePerHost": 1 } }, {

26 VMware, Inc.

"name": "compute", "roles": [ "hadoop_tasktracker" ], "instanceNum": 8, "cpuNum": 2, "memCapacityMB": 7500, "storage": { "type": "LOCAL", "sizeGB": 20 }, "placementPolicies": { "instancePerHost": 2 } }, { "name": "client", "roles": [ "hadoop_client", "hive", "pig" ], "instanceNum": 1, "cpuNum": 1, "storage": { "type": "LOCAL", "sizeGB": 50 } } ], "configuration": { } }

Chapter 3 Creating Hadoop and HBase Clusters

2 Access the Serengeti CLI.

3 Run the cluster create command, and specify the cluster specification file.

cluster create --name cluster_name --specFile full_path/spec_filename

Create a Compute-Only Cluster with the Serengeti Command-Line Interface

You can create compute-only clusters to run MapReduce jobs on existing HDFS clusters, including storage solutions that serve as an external HDFS.

Prerequisites

Deploy the Serengeti vApp.

n

Ensure that you have adequate resources allocated to run the Hadoop cluster.

n

To use any Hadoop distribution other than the provided Apache Hadoop, add one or more Hadoop

n

distributions. See the VMware vSphere Big Data Extensions Administrator's and User's Guide.

Procedure

1 Create a cluster specification file that is modeled on the Serengeti compute_only_cluster.json sample

cluster specification file found in the Serengeti cli/samples directory.

VMware, Inc. 27

VMware vSphere Big Data Extensions Command-Line Interface Guide

2 Add the following code to your new cluster specification file.

For HDFS clusters, set port_num to 8020. For Hadoop 2.0 clusters, such as CDH4 and Pivotal HD distributions, set port_num to 9000.

NOTE To create an Apache Bigtop, Cloudera CDH4 and CDH5, or Pivotal PHD 1.1 cluster, you must configure a valid DNS and FQDN for the cluster's HDFS and MapReduce traffic. Without valid DNS and FQDN settings, the cluster creation process might fail or the cluster is created but does not function.

In this example, the externalHDFS field points to an HDFS. Assign the hadoop_jobtracker role to the

master node group and the hadoop_tasktracker role to the worker node group.

The externalHDFS field conflicts with node groups that have hadoop_namenode and hadoop_datanode roles. This conflict might cause the cluster creation to fail or, if successfully created, the cluster might not work correctly. To avoid this problem, define only a single HDFS.

{ "externalHDFS": "hdfs://hostname-of-namenode:port_num", "nodeGroups": [ { "name": "master", "roles": [ "hadoop_jobtracker" ], "instanceNum": 1, "cpuNum": 2, "memCapacityMB": 7500, }, { "name": "worker", "roles": [ "hadoop_tasktracker", ], "instanceNum": 4, "cpuNum": 2, "memCapacityMB": 7500, "storage": { "type": "LOCAL", "sizeGB": 20 }, }, { "name": "client", "roles": [ "hadoop_client", "hive", "pig" ], "instanceNum": 1, "cpuNum": 1, "storage": { "type": "LOCAL", "sizeGB": 50 }, }

28 VMware, Inc.

Chapter 3 Creating Hadoop and HBase Clusters

], “configuration” : { } }

3 Access the Serengeti CLI.

4 Run the cluster create command and include the cluster specification file parameter and associated

filename.

cluster create --name name_computeOnlyCluster --specFile path/spec_file_name

Create a Basic Cluster with the Serengeti Command-Line Interface

You can create a basic cluster in your Serengeti environment. A basic cluster is a group of virtual machines provisioned and managed by Serengeti. Serengeti helps you to plan and provision the virtual machines to your specifications. You can use the basic cluster's virtual machines to install Big Data applications.

The basic cluster does not install the Big Data application packages used when creating a Hadoop or HBase cluster. Instead, you can install and manage Big Data applications with third party application management tools such as Apache Ambari or Cloudera Manager within your Big Data Extensions environment, and integrate it with your Hadoop software. The basic cluster does not deploy a Hadoop or Hbase cluster. You must deploy software into the basic cluster's virtual machines using an external third party application management tool.

The Serengeti package includes an annotated sample cluster specification file that you can use as an example when you create your basic cluster specification file. In the Serengeti Management Server, the sample specification file is located at /opt/serengeti/samples/basic_cluster.json. You can modify the configuration values in the sample cluster specification file to meet your requirements. The only value you cannot change is the value assigned to the role for each node group, which must always be basic.

You can deploy a basic cluster with the Big Data Extension plug-in using a customized cluster specification file.

To deploy software within the basic cluster virtual machines, use the cluster list --detail command, or run serengeti-ssh.sh cluster_name to obtain the IP address of the virtual machine. You can then use the IP address with management applications such as Apache Ambari or Cloudera Manager to provision the virtual machine with software of your choosing. You can configure the management application to use the user name serengeti, and the password you specified when creating the basic cluster within Big Data Extensions when the management tool needs a user name and password to connect to the virtual machines.

Prerequisites

Deploy the Serengeti vApp.

n

Ensure that you have adequate resources allocated to run the cluster, as well as the Big Data software

n

you intend to deploy.

Procedure

1 Create a specification file to define the basic cluster's characteristics.

You must use the basic role for each node group you define for the basic cluster.

{ "nodeGroups":[ { "name": "master", "roles": [ "basic" ], "instanceNum": 1,

VMware, Inc. 29

VMware vSphere Big Data Extensions Command-Line Interface Guide

"cpuNum": 2, "memCapacityMB": 3768, "storage": { "type": "LOCAL", "sizeGB": 250 }, "haFlag": "on" }, { "name": "worker", "roles": [ "basic" ], "instanceNum": 1, "cpuNum": 2, "memCapacityMB": 3768, "storage": { "type": "LOCAL", "sizeGB": 250 }, "haFlag": "off" } ] }

2 Access the Serengeti CLI.

3 Run the cluster create command, and specify the basic cluster specification file.

cluster create --name cluster_name --specFile /opt/serengeti/samples/basic_cluster.json -password

NOTE When creating a basic cluster, you do not need to specify a Hadoop distribution type using the

--distro option. The reason for this is that there is no Hadoop distribution being installed within the

basic cluster to be managed by Serengeti.

30 VMware, Inc.

About Cluster Topology

You can improve workload balance across your cluster nodes, and improve performance and throughput, by specifying how Hadoop virtual machines are placed using topology awareness. For example, you can have separate data and compute nodes, and improve performance and throughput by placing the nodes on the same set of physical hosts.

To get maximum performance out of your Hadoop or HBase cluster, configure your cluster so that it has awareness of the topology of your environment's host and network information. Hadoop performs better when it uses within-rack transfers, where more bandwidth is available, to off-rack transfers when assigning MapReduce tasks to nodes. HDFS can place replicas more intelligently to trade off performance and resilience. For example, if you have separate data and compute nodes, you can improve performance and throughput by placing the nodes on the same set of physical hosts.

CAUTION When you create a cluster with Big Data Extensions, Big Data Extensions disables the cluster's virtual machine automatic migration. Although this prevents vSphere from migrating the virtual machines, it does not prevent you from inadvertently migrating cluster nodes to other hosts by using the vCenter Server user interface. Do not use the vCenter Server user interface to migrate clusters. Performing such management functions outside of the Big Data Extensions environment might break the cluster's placement policy, such as the number of instances per host and the group associations. Even if you do not specify a placement policy, using vCenter Server to migrate clusters can break the default ROUNDROBIN placement policy constraints.

Chapter 3 Creating Hadoop and HBase Clusters

You can specify the following topology awareness configurations.

Hadoop Virtualization Extensions (HVE)

RACK_AS_RACK

HOST_AS_RACK

None

Enhanced cluster reliability and performance provided by refined Hadoop replica placement, task scheduling, and balancer policies. Hadoop clusters implemented on a virtualized infrastructure have full awareness of the topology on which they are running when using HVE.

To use HVE, your Hadoop distribution must support HVE and you must create and upload a topology rack-hosts mapping file.

Standard topology for Apache Hadoop distributions. Only rack and host information are exposed to Hadoop. To use RACK_AS_RACK, create and upload a server topology file.

Simplified topology for Apache Hadoop distributions. To avoid placing all HDFS data block replicas on the same physical host, each physical host is treated as a rack. Because data block replicas are never placed on a rack, this avoids the worst case scenario of a single host failure causing the complete loss of any data block.

Use HOST_AS_RACK if your cluster uses a single rack, or if you do not have rack information with which to decide about topology configuration options.

No topology is specified.

Topology Rack-Hosts Mapping File

Rack-Hosts mapping files are plain text files that associate logical racks with physical hosts. These files are required to create clusters with HVE or RACK_AS_RACK topology.

The format for every line in a topology rack-hosts mapping file is:

rackname: hostname1, hostname2 ...

VMware, Inc. 31

VMware vSphere Big Data Extensions Command-Line Interface Guide

For example, to assign physical hosts a.b.foo.com and a.c.foo.com to rack1, and physical host c.a.foo.com to rack2, include the following lines in your topology rack-hosts mapping file.

rack1: a.b.foo.com, a.c.foo.com rack2: c.a.foo.com

Topology Placement Policy Definition Files

The placementPolicies field in the cluster specification file controls how nodes are placed in the cluster.

If you specify values for both instancePerHost and groupRacks, there must be a sufficient number of available hosts. To display the rack hosts information, use the topology list command.

The code shows an example placementPolicies field in a cluster specification file.

{ "nodeGroups":[ … { "name": "group_name", … "placementPolicies": { "instancePerHost": 2, "groupRacks": { "type": "ROUNDROBIN", "racks": ["rack1", "rack2", "rack3"] }, "groupAssociations": [{ "reference": "another_group_name", "type": "STRICT" // or "WEAK" }] } }, … }

32 VMware, Inc.

Chapter 3 Creating Hadoop and HBase Clusters

Table 3‑1. placementPolicies Object Definition

JSON field Type Description

instancePerHost Optional Number of virtual machine nodes to

place for each physical ESXi host. This constraint is aimed at balancing the workload.

groupRacks Optional Method of distributing virtual machine

nodes among the cluster’s physical racks. Specify the following JSON strings:

n

type. Specify ROUNDROBIN, which selects candidates fairly and without priority.

n

racks. Which racks in the topology map to use.

groupAssociations Optional One or more target node groups with

which this node group associates. Specify the following JSON strings:

n

reference. Target node group name

n

type:

STRICT. Place the node group on

n

the target group’s set or subset of ESXi hosts. If STRICT placement is not possible, the operation fails.

WEAK. Attempt to place the node

n

group on the target group’s set or subset of ESXi hosts, but if that is not possible, use an extra ESXi host.

Create a Cluster with Topology Awareness with the Serengeti Command-Line Interface

To achieve a balanced workload or to improve performance and throughput, you can control how Hadoop virtual machines are placed by adding topology awareness to the Hadoop clusters. For example, you can have separate data and compute nodes, and improve performance and throughput by placing the nodes on the same set of physical hosts.

Prerequisites

Deploy the Serengeti vApp.

n

Ensure that you have adequate resources allocated to run the Hadoop cluster.

n

To use any Hadoop distribution other than the provided Apache Hadoop, add one or more Hadoop

n

distributions. See the VMware vSphere Big Data Extensions Administrator's and User's Guide.

Procedure

1 Access the Serengeti Command-Line Interface.

2 (Optional) Run the topology list command to view the list of available topologies.

topology list

3 (Optional) If you want the cluster to use HVE or RACK_AS_RACK toplogies, create a topology rack-

hosts mapping file and upload the file to the Serengeti Management Server.

topology upload --fileName name_of_rack_hosts_mapping_file

VMware, Inc. 33

VMware vSphere Big Data Extensions Command-Line Interface Guide

4 Run the cluster create command to create the cluster.

cluster create --name cluster-name ... --topology {HVE|RACK_AS_RACK|HOST_AS_RACK}

NOTE To create an Apache Bigtop, Cloudera CDH4 and CDH5, or Pivotal PHD 1.1 cluster, you must configure a valid DNS and FQDN for the cluster's HDFS and MapReduce traffic. Without valid DNS and FQDN settings, the cluster creation process might fail or the cluster is created but does not function.

This example creates an HVE topology.

cluster create --name cluster-name --topology HVE --distro name_of_HVE-supported_distro

5 View the allocated nodes on each rack.

cluster list --name cluster-name –-detail

Create a Data-Compute Separated Cluster with Topology Awareness and Placement Constraints

You can create clusters with separate data and compute nodes, and define topology and placement policy constraints to distribute the nodes among the physical racks and the virtual machines.

CAUTION When you create a cluster with Big Data Extensions, Big Data Extensions disables the cluster's virtual machine automatic migration. Although this prevents vSphere from migrating the virtual machines, it does not prevent you from inadvertently migrating cluster nodes to other hosts by using the vCenter Server user interface. Do not use the vCenter Server user interface to migrate clusters. Performing such management functions outside of the Big Data Extensions environment might break the cluster's placement policy, such as the number of instances per host and the group associations. Even if you do not specify a placement policy, using vCenter Server to migrate clusters can break the default ROUNDROBIN placement policy constraints.

Prerequisites

Deploy the Serengeti vApp.

n

Ensure that you have adequate resources allocated to run the Hadoop cluster.

n

To use any Hadoop distribution other than the provided Apache Hadoop, add one or more Hadoop

n

distributions. See the VMware vSphere Big Data Extensions Administrator's and User's Guide.

Create a rack-host mapping information file.

n

Upload the rack-host file to the Serengeti server with the topology upload command.

n

Procedure

1 Create a cluster specification file to define the cluster's characteristics, including the node groups,

topology, and placement constraints.

NOTE To create an Apache Bigtop, Cloudera CDH4 and CDH5, or Pivotal PHD 1.1 cluster, you must configure a valid DNS and FQDN for the cluster's HDFS and MapReduce traffic. Without valid DNS and FQDN settings, the cluster creation process might fail or the cluster is created but does not function.

In this example, the cluster has groupAssociations and instancePerHost constraints for the compute node group, and a groupRacks constraint for the data node group.

34 VMware, Inc.

Chapter 3 Creating Hadoop and HBase Clusters

Four data nodes and eight compute nodes are placed on the same four ESXi hosts, which are fairly selected from rack1, rack2, and rack3. Each ESXi host has one data node and two compute nodes. As defined for the compute node group, compute nodes are placed only on ESXi hosts that have data nodes.

This cluster definition requires that you configure datastores and resource pools for at least four hosts, and that there is sufficient disk space for Serengeti to perform the necessary placements during deployment.

{ "nodeGroups":[ { "name": "master", "roles": [ "hadoop_namenode", "hadoop_jobtracker" ], "instanceNum": 1, "cpuNum": 2, "memCapacityMB": 7500, }, { "name": "data", "roles": [ "hadoop_datanode" ], "instanceNum": 4, "cpuNum": 1, "memCapacityMB": 3748, "storage": { "type": "LOCAL", "sizeGB": 50 }, "placementPolicies": { "instancePerHost": 1, "groupRacks": { "type": "ROUNDROBIN", "racks": ["rack1", "rack2", "rack3"] }, } }, { "name": "compute", "roles": [ "hadoop_tasktracker" ], "instanceNum": 8, "cpuNum": 2, "memCapacityMB": 7500, "storage": { "type": "LOCAL", "sizeGB": 20 }, "placementPolicies": { "instancePerHost": 2, "groupAssociations": [

VMware, Inc. 35

VMware vSphere Big Data Extensions Command-Line Interface Guide

{ "reference": "data", "type": "STRICT" } } }, { "name": "client", "roles": [ "hadoop_client", "hive", "pig" ], "instanceNum": 1, "cpuNum": 1, "storage": { "type": "LOCAL", "sizeGB": 50 } } ], "configuration": { } }

2 Access the Serengeti CLI.

3 Run the cluster create command, and specify the cluster specification file.

cluster create --name cluster_name --specFile full_path/spec_filename

Serengeti’s Default HBase Cluster Configuration

HBase clusters are required for you to build big table applications. To run HBase MapReduce jobs, configure the HBase cluster to include JobTracker nodes or TaskTracker nodes.

When you create an HBase cluster with the Command-Line Interface, according to the default Serengeti HBase template the resulting cluster consists of the following nodes:

One master node, which runs the NameNode and HBaseMaster services.

n

Three zookeeper nodes, each running the ZooKeeper service.

n

Three data nodes, each running the DataNode and HBase Regionserver services.

n

One client node, from which you can run Hadoop or HBase jobs.

n

The default Serengeti-deployed HBase cluster does not contain Hadoop JobTracker or Hadoop TaskTracker daemons. To run an HBase MapReduce job, deploy a customized, nondefault HBase cluster.

Create a Default HBase Cluster with the Serengeti Command-Line Interface

Serengeti supports deploying HBase clusters on HDFS.

Prerequisites

Deploy the Serengeti vApp.

n

Ensure that you have adequate resources allocated to run the Hadoop cluster.

n

36 VMware, Inc.

Chapter 3 Creating Hadoop and HBase Clusters

To use any Hadoop distribution other than the provided Apache Hadoop, add one or more Hadoop

n

distributions. See the VMware vSphere Big Data Extensions Administrator's and User's Guide.

Procedure

1 Access the Serengeti CLI.

2 Run the cluster create command, and specify the --type parameter’s value as hbase.

cluster create --name cluster_name --type hbase

What to do next

After you deploy the cluster, you can access an HBase database by using several methods. See the VMware vSphere Big Data Extensions Administrator's and User's Guide.

Create an HBase Cluster with vSphere HA Protection with the Serengeti Command-Line Interface

You can create HBase clusters with separated Hadoop NameNode and HBase Master roles, and configure vSphere HA protection for the Masters.

Prerequisites

Deploy the Serengeti vApp.

n

Ensure that you have adequate resources allocated to run the Hadoop cluster.

n

To use any Hadoop distribution other than the provided Apache Hadoop, add one or more Hadoop

n

distributions. See the VMware vSphere Big Data Extensions Administrator's and User's Guide.

Procedure

1 Create a cluster specification file to define the cluster's characteristics, including the node group roles

and vSphere HA protection.

In this example, the cluster has JobTracker and TaskTracker nodes, which let you run HBase MapReduce jobs. The Hadoop NameNode and HBase Master roles are separated, and both are protected by vSphere HA.

{ "nodeGroups" : [ { "name" : "zookeeper", "roles" : [ "zookeeper" ], "instanceNum" : 3, "instanceType" : "SMALL", "storage" : { "type" : "shared", "sizeGB" : 20 }, "cpuNum" : 1, "memCapacityMB" : 3748, "haFlag" : "on", "configuration" : { } }, { "name" : "hadoopmaster",

VMware, Inc. 37

VMware vSphere Big Data Extensions Command-Line Interface Guide

"roles" : [ "hadoop_namenode", "hadoop_jobtracker" ], "instanceNum" : 1, "instanceType" : "MEDIUM", "storage" : { "type" : "shared", "sizeGB" : 50 }, "cpuNum" : 2, "memCapacityMB" : 7500, "haFlag" : "on", "configuration" : { } }, { "name" : "hbasemaster", "roles" : [ "hbase_master" ], "instanceNum" : 1, "instanceType" : "MEDIUM", "storage" : { "type" : "shared", "sizeGB" : 50 }, "cpuNum" : 2, "memCapacityMB" : 7500, "haFlag" : "on", "configuration" : { } },

{ "name" : "worker", "roles" : [ "hadoop_datanode", "hadoop_tasktracker", "hbase_regionserver" ], "instanceNum" : 3, "instanceType" : "SMALL", "storage" : { "type" : "local", "sizeGB" : 50 }, "cpuNum" : 1, "memCapacityMB" : 3748, "haFlag" : "off", "configuration" : { } }, { "name" : "client",

38 VMware, Inc.

Chapter 3 Creating Hadoop and HBase Clusters

"roles" : [ "hadoop_client", "hbase_client" ], "instanceNum" : 1, "instanceType" : "SMALL", "storage" : { "type" : "shared", "sizeGB" : 50 }, "cpuNum" : 1, "memCapacityMB" : 3748, "haFlag" : "off", "configuration" : { } } ], // we suggest running convert-hadoop-conf.rb to generate "configuration" section and paste the output here "configuration" : { "hadoop": { "core-site.xml": { // check for all settings at http://hadoop.apache.org/common/docs/stable/coredefault.html // note: any value (int, float, boolean, string) must be enclosed in double quotes and here is a sample: // "io.file.buffer.size": "4096" }, "hdfs-site.xml": { // check for all settings at http://hadoop.apache.org/common/docs/stable/hdfsdefault.html }, "mapred-site.xml": { // check for all settings at http://hadoop.apache.org/common/docs/stable/mapreddefault.html }, "hadoop-env.sh": { // "HADOOP_HEAPSIZE": "", // "HADOOP_NAMENODE_OPTS": "", // "HADOOP_DATANODE_OPTS": "", // "HADOOP_SECONDARYNAMENODE_OPTS": "", // "HADOOP_JOBTRACKER_OPTS": "", // "HADOOP_TASKTRACKER_OPTS": "", // "HADOOP_CLASSPATH": "", // "JAVA_HOME": "", // "PATH": "" }, "log4j.properties": { // "hadoop.root.logger": "DEBUG,DRFA", // "hadoop.security.logger": "DEBUG,DRFA" }, "fair-scheduler.xml": { // check for all settings at http://hadoop.apache.org/docs/stable/fair_scheduler.html // "text": "the full content of fair-scheduler.xml in one line"

VMware, Inc. 39

VMware vSphere Big Data Extensions Command-Line Interface Guide

}, "capacity-scheduler.xml": { // check for all settings at http://hadoop.apache.org/docs/stable/capacity_scheduler.html }, "mapred-queue-acls.xml": { // check for all settings at http://hadoop.apache.org/docs/stable/cluster_setup.html#Configuring+the+Hadoop+Daemons // "mapred.queue.queue-name.acl-submit-job": "", // "mapred.queue.queue-name.acl-administer-jobs", "" } }, "hbase": { "hbase-site.xml": { // check for all settings at http://hbase.apache.org/configuration.html#hbase.site }, "hbase-env.sh": { // "JAVA_HOME": "", // "PATH": "", // "HBASE_CLASSPATH": "", // "HBASE_HEAPSIZE": "", // "HBASE_OPTS": "", // "HBASE_USE_GC_LOGFILE": "", // "HBASE_JMX_BASE": "", // "HBASE_MASTER_OPTS": "", // "HBASE_REGIONSERVER_OPTS": "", // "HBASE_THRIFT_OPTS": "", // "HBASE_ZOOKEEPER_OPTS": "", // "HBASE_REGIONSERVERS": "", // "HBASE_SSH_OPTS": "", // "HBASE_NICENESS": "", // "HBASE_SLAVE_SLEEP": "" }, "log4j.properties": { // "hbase.root.logger": "DEBUG,DRFA" } }, "zookeeper": { "java.env": { // "JVMFLAGS": "-Xmx2g" }, "log4j.properties": { // "zookeeper.root.logger": "DEBUG,DRFA" } } } }

2 Access the Serengeti CLI.

3 Run the cluster create command, and specify the cluster specification file.

cluster create --name cluster_name --specFile full_path/spec_filename

40 VMware, Inc.

Managing Hadoop and HBase

Clusters 4

You can use the vSphere Web Client to start and stop your Hadoop or HBase cluster and to modify cluster configuration. You can also manage a cluster using the Serengeti Command-Line Interface.

CAUTION Do not use vSphere management functions such as migrating cluster nodes to other hosts for clusters that you create with Big Data Extensions. Performing such management functions outside of the Big Data Extensions environment can make it impossible for you to perform some Big Data Extensions operations, such as disk failure recovery.

Stop and Start a Hadoop or HBase Cluster with the Serengeti Command-Line Interface on page 42

n

You can stop a currently running cluster and start a stopped cluster from the Serengeti CLI.

Scale Out a Hadoop or HBase Cluster with the Serengeti Command-Line Interface on page 42

n

You specify the number of nodes in the cluster when you create Hadoop and HBase clusters. You can later scale out the cluster by increasing the number of worker nodes and client nodes.

Scale CPU and RAM with the Serengeti Command-Line Interface on page 43

n

You can increase or decrease a Hadoop or HBase cluster’s compute capacity and RAM to prevent memory resource contention of running jobs.

Reconfigure a Hadoop or HBase Cluster with the Serengeti Command-Line Interface on page 43

n

You can reconfigure any Hadoop or HBase cluster that you create with Big Data Extensions.

About Resource Usage and Elastic Scaling on page 45

n

Scaling lets you adjust the compute capacity of Hadoop data-compute separated clusters. When you enable elastic scaling for a Hadoop cluster, the Serengeti Management Server can stop and start compute nodes to match resource requirements to available resources. You can use manual scaling for more explicit cluster control.

Delete a Hadoop or HBase Cluster with the Serengeti Command-Line Interface on page 51

n

You can delete a Hadoop cluster that you no longer need, regardless of whether it is running. When a Hadoop cluster is deleted, all its virtual machines and resource pools are destroyed.

About vSphere High Availability and vSphere Fault Tolerance on page 51

n

The Serengeti Management Server leverages vSphere HA to protect the Hadoop master node virtual machine, which can be monitored by vSphere.

Reconfigure a Node Group with the Serengeti Command-Line Interface on page 51

n

You can reconfigure node groups by modifying node group configuration data in the associated cluster specification file. When you configure a node group, its configuration overrides any cluster level configuration of the same name.

VMware, Inc.

41

VMware vSphere Big Data Extensions Command-Line Interface Guide

Recover from Disk Failure with the Serengeti Command-Line Interface Client on page 51

n

If there is a disk failure in a Hadoop cluster, and the disk does not perform management roles such as NameNode, JobTracker, ResourceManager, HMaster, or ZooKeeper, you can recover by running the Serengeti cluster fix command.

Stop and Start a Hadoop or HBase Cluster with the Serengeti Command-Line Interface

You can stop a currently running cluster and start a stopped cluster from the Serengeti CLI.

Prerequisites

Verify that the cluster is provisioned.

n

Verify that enough resources, especially CPU and memory, are available to start the virtual machines in

n

the Hadoop cluster.

Procedure

1 Access the Serengeti CLI.

2 Run the cluster stop command.

cluster stop –-name name_of_cluster_to_stop

3 Run the cluster start command.

cluster start –-name name_of_cluster_to_start

Scale Out a Hadoop or HBase Cluster with the Serengeti CommandLine Interface

You specify the number of nodes in the cluster when you create Hadoop and HBase clusters. You can later scale out the cluster by increasing the number of worker nodes and client nodes.

IMPORTANT Even if you changed the user password on the cluster's nodes, the changed password is not used for the new nodes that are created when you scale out a cluster. If you set the cluster's initial administrator password when you created the cluster, that initial administrator password is used for the new nodes. If you did not set the cluster's initial administrator password when you created the cluster, new random passwords are used for the new nodes.

Prerequisites

If the cluster is stopped, start it. The cluster status must be RUNNING.

Procedure

1 Access the Serengeti CLI.

2 Run the cluster resize command.

For node_type, specify worker or client. For the instanceNum parameter’s num_nodes value, use any number that is larger than the current number of node_type instances.

cluster resize --name name_of_cluster_to_resize --nodeGroup node_type --instanceNum num_nodes

42 VMware, Inc.

Chapter 4 Managing Hadoop and HBase Clusters

Scale CPU and RAM with the Serengeti Command-Line Interface

You can increase or decrease a Hadoop or HBase cluster’s compute capacity and RAM to prevent memory resource contention of running jobs.

Serengeti lets you adjust compute and memory resources without increasing the workload on the master node. If increasing or decreasing the cluster's CPU is unsuccessful for a node, which is commonly due to insufficient resources being available, the node is returned to its original CPU setting. If increasing or decreasing the cluster's RAM is unsuccessful for a node, which is commonly due to insufficient resources, the swap disk retains its new setting anyway. The disk is not returned to its original memory setting.

Although all node types support CPU and RAM scaling, do not scale a cluster's master node because Serengeti powers down the virtual machine during the scaling process.

The maximum CPU and RAM settings depend on the virtual machine's version.

Table 4‑1. Maximum CPU and RAM Settings

Virtual Machine Version Maximum Number of CPUs Maximum RAM, in GB

7 8 255

8 32 1011

9 64 1011

10 64 1011

Prerequisites

Start the cluster if it is not running.

Procedure

1 Access the Serengeti Command-Line Interface.

2 Run the cluster resize command to change the number of CPUs or the amount of RAM of a cluster.

Node types are either worker or client.

n

Specify one or both scaling parameters: --cpuNumPerNode or --memCapacityMbPerNode.

n

cluster resize --name cluster_name --nodeGroup node_type [--cpuNumPerNode vCPUs_per_node] [-memCapacityMbPerNode memory_per_node]

Reconfigure a Hadoop or HBase Cluster with the Serengeti Command-Line Interface

You can reconfigure any Hadoop or HBase cluster that you create with Big Data Extensions.

The cluster configuration is specified by attributes in Hadoop distribution XML configuration files such as:

core-site.xml, hdfs-site.xml, mapred-site.xml, hadoop-env.sh, yarn-env.sh, yarn-site.sh, and hadoopmetrics.properties.

NOTE Always use the cluster config command to change the parameters specified by these configuration files. If you manually modify these files, your changes will be erased if the virtual machine is rebooted, or you use the cluster config, cluster start, cluster stop, or cluster resize commands.

VMware, Inc. 43

VMware vSphere Big Data Extensions Command-Line Interface Guide

Procedure

1 Use the cluster export command to export the cluster specification file for the cluster that you want to

reconfigure.

cluster export --name cluster_name --specFile file_path/cluster_spec_file_name

Option Description

cluster_name

file_path

cluster_spec_file_name

Name of the cluster that you want to reconfigure.

The file system path at which to export the specification file.

The name with which to label the exported cluster specification file.

2 Edit the configuration information located near the end of the exported cluster specification file.

If you are modeling your configuration file on existing Hadoop XML configuration files, use the

convert-hadoop-conf.rb conversion tool to convert Hadoop XML configuration files to the required

JSON format.

… "configuration": { "hadoop": { "core-site.xml": { // check for all settings at http://hadoop.apache.org/common/docs/stable/coredefault.html // note: any value (int, float, boolean, string) must be enclosed in double quotes and here is a sample: // "io.file.buffer.size": "4096" }, "hdfs-site.xml": { // check for all settings at http://hadoop.apache.org/common/docs/stable/hdfsdefault.html }, "mapred-site.xml": { // check for all settings at http://hadoop.apache.org/common/docs/stable/mapreddefault.html }, "hadoop-env.sh": { // "HADOOP_HEAPSIZE": "", // "HADOOP_NAMENODE_OPTS": "", // "HADOOP_DATANODE_OPTS": "", // "HADOOP_SECONDARYNAMENODE_OPTS": "", // "HADOOP_JOBTRACKER_OPTS": "", // "HADOOP_TASKTRACKER_OPTS": "", // "HADOOP_CLASSPATH": "", // "JAVA_HOME": "", // "PATH": "", }, "log4j.properties": { // "hadoop.root.logger": "DEBUG, DRFA ", // "hadoop.security.logger": "DEBUG, DRFA ", }, "fair-scheduler.xml": { // check for all settings at http://hadoop.apache.org/docs/stable/fair_scheduler.html // "text": "the full content of fair-scheduler.xml in one line" },

44 VMware, Inc.

Chapter 4 Managing Hadoop and HBase Clusters

"capacity-scheduler.xml": { // check for all settings at http://hadoop.apache.org/docs/stable/capacity_scheduler.html } } } …

3 (Optional) If your Hadoop distribution’s JAR files are not in the $HADOOP_HOME/lib directory, add the

full path of the JAR file in $HADOOP_CLASSPATH to the cluster specification file.

This action lets the Hadoop daemons locate the distribution JAR files.

For example, the Cloudera CDH3 Hadoop Fair Scheduler JAR files are in /usr/lib/hadoop/contrib/fairscheduler/. Add the following to the cluster specification file to enable Hadoop to use the JAR files.

… "configuration": { "hadoop": { "hadoop-env.sh": { "HADOOP_CLASSPATH": "/usr/lib/hadoop/contrib/fairscheduler/*:$HADOOP_CLASSPATH" }, "mapred-site.xml": { "mapred.jobtracker.taskScheduler": "org.apache.hadoop.mapred.FairScheduler" … }, "fair-scheduler.xml": { … } } } …

4 Access the Serengeti Command-Line Interface.

5 Run the cluster config command to apply the new Hadoop configuration.

cluster config --name cluster_name --specFile file_path/cluster_spec_file_name

6 (Optional) Reset an existing configuration attribute to its default value.

a Remove the attribute from the cluster configuration file’s configuration section, or comment out the

attribute using double back slashes (//).

b Re-run the cluster config command.

About Resource Usage and Elastic Scaling

Scaling lets you adjust the compute capacity of Hadoop data-compute separated clusters. When you enable elastic scaling for a Hadoop cluster, the Serengeti Management Server can stop and start compute nodes to match resource requirements to available resources. You can use manual scaling for more explicit cluster control.

Manual scaling is appropriate for static environments where capacity planning can predict resource availability for workloads. Elastic scaling is best suited for mixed workload environments where resource requirements and availability fluctuate.

VMware, Inc. 45

VMware vSphere Big Data Extensions Command-Line Interface Guide

When you select manual scaling, Big Data Extensions disables elastic scaling. You can configure the target number of compute nodes for manual scaling. If you do not configure the target number of compute nodes, Big Data Extensions sets the number of active compute nodes to the current number of active compute nodes. If nodes become unresponsive, they remain in the cluster and the cluster operates with fewer functional nodes. In contrast, when you enable elastic scaling, Big Data Extensions manages the number of active TaskTracker nodes according to the range that you specify, replacing unresponsive or faulty nodes with live, responsive nodes.

For both manual and elastic scaling, Big Data Extensions, not vCenter Server, controls the number of active nodes. However, vCenter Server applies the usual reservations, shares, and limits to the cluster's resource pool according to the cluster's vSphere configuration. vSphere DRS operates as usual, allocating resources between competing workloads, which in turn influences how Big Data Extensions dynamically adjusts the number of active nodes in competing Hadoop clusters while elastic scaling is in effect.

Big Data Extensions also lets you adjust cluster nodes' access priority for datastores by using the vSphere Storage I/O Control feature. Clusters configured for HIGH I/O shares receive higher priority access than clusters with NORMAL priority. Clusters configured for NORMAL I/O shares receive higher priority access than clusters with LOW priority. In general, higher priority provides better disk I/O performance.

Scaling Modes

To change between manual and elastic scaling, you change the scaling mode.

MANUAL. Big Data Extensions disables elastic scaling. When you change to manual scaling, you can

n

configure the target number of compute nodes. If you do not configure the target number of compute nodes, Big Data Extensions sets the number of active compute nodes to the current number of active compute nodes.

AUTO. Enables elastic scaling. Big Data Extensions manages the number of active compute nodes,

n

maintaining the number of compute nodes in the range from the configured minimum to the configured maximum number of compute nodes in the cluster. If the minimum number of compute nodes is undefined, the lower limit is 0. If the maximum number of compute nodes is undefined, the upper limit is the number of available compute nodes.

Elastic scaling operates on a per-host basis, at a node-level granularity. That is, the more compute nodes a Hadoop cluster has on a host, the finer the control that Big Data Extensions elasticity can exercise. The tradeoff is that the more compute nodes you have, the higher the overhead in terms of runtime resource cost, disk footprint, I/O requirements, and so on.

When resources are overcommitted, elastic scaling reduces the number of powered on compute nodes. Conversely, if the cluster receives all the resources it requested from vSphere, and Big Data Extensions determines that the cluster can make use of additional capacity, elastic scaling powers on additional compute nodes.

Resources can become overcommitted for many reasons, such as:

The compute nodes have lower resource entitlements than a competing workload, according to

n

how vCenter Server applies the usual reservations, shares, and limits as configured for the cluster.

Physical resources are configured to be available, but another workload is consuming those

n

resources.

In elastic scaling, Big Data Extensions has two different behaviors for deciding how many active compute nodes to maintain. In both behaviors, Big Data Extensions replaces unresponsive or faulty nodes with live, responsive nodes.

Variable. The number of active, healthy TaskTracker compute nodes is maintained from the

n

configured minimum number of compute nodes to the configured maximum number of compute nodes. The number of active compute nodes varies as resource availability fluctuates.

Fixed. The number of active, healthy TaskTracker compute nodes is maintained at a fixed number

n

when the same value is configured for the minimum and maximum number of compute nodes.

46 VMware, Inc.

Chapter 4 Managing Hadoop and HBase Clusters

Default Cluster Scaling Parameter Values

When you create a cluster, its scaling configuration is as follows.

The cluster's scaling mode is MANUAL, for manual scaling.

n

The cluster's minimum number of compute nodes is -1. It appears as "Unset" in the Serengeti CLI

n

displays. Big Data Extensions elastic scaling treats a minComputeNodeNum value of -1 as if it were zero (0).

The cluster's maximum number of compute nodes is -1. It appears as "Unset" in the Serengeti CLI

n

displays. Big Data Extensions elastic scaling treats a maxComputeNodeNum value of -1 as if it were unlimited.

The cluster's target number of nodes is not applicable. Its value is -1. Big Data Extensions manual

n

scaling operations treat a targetComputeNodeNum value of -1 as if it were unspecified upon a change to manual scaling.

Interactions Between Scaling and Other Cluster Operations

Some cluster operations cannot be performed while Big Data Extensions is actively scaling a cluster.

If you try to perform the following operations while Big Data Extensions is scaling a cluster in MANUAL mode, Big Data Extensions warns you that in the cluster's current state, the operation cannot be performed.

Concurrent attempt at manual scaling

n

Switch to AUTO mode while manual scaling operations are in progress

n

If a cluster is in AUTO mode for elastic scaling when you perform the following cluster operations on it, Big Data Extensions changes the scaling mode to MANUAL and changes the cluster to manual scaling. You can re-enable the AUTO mode for elastic scaling after the cluster operation finishes, except if you delete the cluster.

Delete the cluster

n

Repair the cluster

n

Stop the cluster

n

If a cluster is in AUTO mode for elastic scaling when you perform the following cluster operations on it, Big Data Extensions temporarily switches the cluster to MANUAL mode. When the cluster operation finishes, Big Data Extensions returns the scaling mode to AUTO, which re-enables elastic scaling.

Resize the cluster

n

Reconfigure the cluster

n

If Big Data Extensions is scaling a cluster when you perform an operation that changes the scaling mode to MANUAL, your requested operation waits until the scaling finishes, and then the requested operation begins.

Enable Elastic Scaling for a Hadoop Cluster with the Serengeti Command-Line Interface

When you enable elastic scaling for a data-compute separated Hadoop cluster, Big Data Extensions optimizes cluster performance and utilization of TaskTracker compute nodes.

To enable elastic scaling, set a data-compute separated Hadoop cluster's scaling mode to AUTO and configure the minimum and maximum number of compute nodes. If you do not configure the minimum or maximum number of compute nodes, the previous minimum or maximum setting, respectively, is retained.

VMware, Inc. 47

VMware vSphere Big Data Extensions Command-Line Interface Guide

To ensure that under contention, elastic scaling keeps a cluster operating with more than a cluster’s initial default setting of zero compute nodes, configure the minComputeNodeNum parameter value to a nonzero number. To limit the maximum number of compute nodes that can be used in a Hadoop cluster, configure the maxComputeNodeNum parameter value to less than the total available compute nodes.

In elastic scaling, Big Data Extensions has two different behaviors for deciding how many active compute nodes to maintain. In both behaviors, Big Data Extensions replaces unresponsive or faulty nodes with live, responsive nodes.

Variable. The number of active, healthy TaskTracker compute nodes is maintained from the configured

n

minimum number of compute nodes to the configured maximum number of compute nodes. The number of active compute nodes varies as resource availability fluctuates.

Fixed. The number of active, healthy TaskTracker compute nodes is maintained at a fixed number when

n

the same value is configured for the minimum and maximum number of compute nodes.

Prerequisites

Understand how elastic scaling and resource usage work. See “About Resource Usage and Elastic

n

Scaling,” on page 45.

Verify that the cluster you want to optimize is data-compute separated. See “About Hadoop and HBase

n

Cluster Deployment Types,” on page 17.

Procedure

1 Access the Serengeti CLI.

2 Run the cluster setParam command, and set the --elasticityMode parameter value to AUTO.

cluster setParam --name cluster_name --elasticityMode AUTO [--minComputeNodeNum minNum] [-maxComputeNodeNum maxNum]

Enable Manual Scaling for a Hadoop Cluster with the Serengeti Command-Line Interface

When you enable manual scaling for a cluster, Big Data Extensions disables elastic scaling. When you enable manual scaling, you can configure the target number of compute nodes. If you do not configure the target number of compute nodes, Big Data Extensions sets the number of active compute nodes to the current number of active compute nodes.

Procedure

1 Access the Serengeti CLI.

2 Run the cluster setParam command, and set the --elasticityMode parameter value to MANUAL.

cluster setParam --name cluster_name --elasticityMode MANUAL [--targetComputeNodeNum numTargetNodes]

Configure Scaling Parameters with the Serengeti Command-Line Interface

You can configure scaling parameters, such as the target number of nodes, with or without changing the scaling mode.

Procedure

1 Access the Serengeti CLI.

2 To display a cluster's scaling settings, run the cluster list command.

cluster list --detail --name cluster_name

48 VMware, Inc.

Chapter 4 Managing Hadoop and HBase Clusters

3 To configure one or more scaling parameters, run the cluster setParam command.

The --name parameter is required, and you can include as few or as many of the other parameters as you want. You can repeatedly run the command to configure or reconfigure additional scaling parameters.

cluster setParam –-name cluster_name --elasticityMode mode --targetComputeNodeNum numTargetNodes --minComputeNodeNum minNum --maxComputeNodeNum maxNum --ioShares level

Paremeter Option Description

cluster_name

mode

numTargetNodes

minNum

maxNum

level

Name of the cluster. Specify this parameter every time you run the cluster setParam command.

MANUAL or AUTO.

Number of nodes. This parameter is applicable only for MANUAL scaling mode.

Lower limit of the range of active compute nodes to maintain in the cluster. This parameter is applicable only for AUTO scaling mode.

Upper limit of the range of active compute nodes to maintain in the cluster. This parameter is applicable only for AUTO scaling mode.

LOW, NORMAL, or HIGH.

4 To reset one or more scaling parameters to their default values, run the cluster resetParam command.

The --name parameter is required, and you can include as few or as many of the other parameters as you want. You can repeatedly run the command to reset additional scaling parameters.

For data-compute separated nodes, you can reset all the scaling parameters to their defaults by using the --all parameter.

cluster resetParam -–name cluster_name [--all] [--elasticityMode] [--targetComputeNodeNum] [--minComputeNodeNum] [--maxComputeNodeNum] [--ioShares]

Parameter Description

cluster_name

--all

--elasticityMode

--targetComputeNodeNum

--minComputeNodeNum

--maxComputeNodeNum

--ioShares

Name of the cluster. Specify this parameter every time you run the cluster resetParam command.

Reset all scaling parameters to their defaults.

Sets the scaling mode to MANUAL.

Reset targetComputeNodeNum to -1. Big Data Extensions manual scaling operations treat a targetComputeNodeNum value of -1 as if it were unspecified upon a change to manual scaling.

Reset minComputeNodeNum to 0. It appears as "Unset" in the Serengeti CLI displays. Big Data Extensions elastic scaling treats a minComputeNodeNum value of -1 as if it were zero (0).

Reset maxComputeNodeNum to unlimited. It appears as "Unset" in the Serengeti CLI displays. Big Data Extensions elastic scaling treats a maxComputeNodeNum value of -1 as if it were unlimited.

Reset ioShares to NORMAL.

VMware, Inc. 49

VMware vSphere Big Data Extensions Command-Line Interface Guide

Schedule Fixed Elastic Scaling for a Hadoop Cluster

You can enable fixed, elastic scaling according to a preconfigured schedule. Scheduled fixed, elastic scaling provides more control than variable, elastic scaling while still improving efficiency, allowing explicit changes in the number of active compute nodes during periods of predictable usage.

For example, in an office with typical workday hours, there is likely a reduced load on a VMware View resource pool after the office staff goes home. You could configure scheduled fixed, elastic scaling to specify a greater number of compute nodes from 8 PM to 4 AM, when you know that the workload would otherwise be very light.

Prerequisites

From the Serengeti Command-Line Interface, enable the cluster for elastic scaling, and set the

minComputeNodeNum and MaxComputeNodeNum parameters to the same value: the number of active TaskTracker

nodes that you want during the period of scheduled fixed elasticity.

Procedure

1 Open a command shell, such as Bash or PuTTY, and log in to the Serengeti Management Server as user

serengeti.

2 Use any scheduling mechanism that you want to call

the /opt/serengeti/sbin/set_compute_node_num.sh script to set the number of active TaskTracker compute nodes that you want.

/opt/serengeti/sbin/set_compute_node_num.sh --name cluster_name --computeNodeNum num_TT_to_maintain

After the scheduling mechanism calls the set_compute_node_num.sh script, fixed, elastic scaling remains in effect with the configured number of active TaskTracker compute nodes until the next scheduling mechanism change or until a user changes the scaling mode or parameters in either the vSphere Web Client or the Serengeti Command-Line Interface.

This example shows how to use a crontab file on the Serengeti Management Server to schedule specific numbers of active TaskTracker compute nodes.

# cluster_A: use 20 active TaskTracker compute nodes from 11:00 to 16:00, and 30 compute nodes the rest of the day 00 11 * * * /opt/serengeti/sbin/set_compute_node_num.sh --name cluster_A -computeNodeNum 20 >> $HOME/schedule_elasticity.log 2>&1 00 16 * * * /opt/serengeti/sbin/set_compute_node_num.sh --name cluster_A -computeNodeNum 30 >> $HOME/schedule_elasticity.log 2>&1

# cluster_B: use 3 active TaskTracker compute nodes beginning at 10:00 every weekday 0 10 * * 1-5 /opt/serengeti/sbin/set_compute_node_num.sh --name cluster_B -computeNodeNum 3 >> $HOME/schedule_elasticity.log 2>&1

# cluster_C: reset the number of active TaskTracker compute nodes every 6 hours to 15 0 */6 * * * /opt/serengeti/sbin/set_compute_node_num.sh --name cluster_B -computeNodeNum 15 >> $HOME/schedule_elasticity.log 2>&1

50 VMware, Inc.

Chapter 4 Managing Hadoop and HBase Clusters

Delete a Hadoop or HBase Cluster with the Serengeti Command-Line Interface

You can delete a Hadoop cluster that you no longer need, regardless of whether it is running. When a Hadoop cluster is deleted, all its virtual machines and resource pools are destroyed.

Procedure

1 Access the Serengeti CLI.

2 Run the cluster delete command.

cluster delete --name cluster_name

About vSphere High Availability and vSphere Fault Tolerance

The Serengeti Management Server leverages vSphere HA to protect the Hadoop master node virtual machine, which can be monitored by vSphere.

When a Hadoop NameNode or JobTracker service stops unexpectedly, vSphere restarts the Hadoop virtual machine in another host, reducing unplanned downtime. If vsphere Fault Tolerance is configured and the master node virtual machine stops unexpectedly because of host failover or loss of network connectivity, the secondary node is used, without downtime.

Reconfigure a Node Group with the Serengeti Command-Line Interface

You can reconfigure node groups by modifying node group configuration data in the associated cluster specification file. When you configure a node group, its configuration overrides any cluster level configuration of the same name.

Procedure

1 Access the Serengeti CLI.

2 Run the cluster export command to export the cluster’s cluster specification file.

cluster export --name cluster_name --specFile path_name/spec_file_name

3 In the specification file, modify the node group’s configuration section with the same content as a

cluster-level configuration.

4 Add the customized Hadoop configuration for the node group that you want to reconfigure.

5 Run the cluster config command to apply the new Hadoop configuration.

cluster config --name cluster_name --specFile path_name/spec_file_name

Recover from Disk Failure with the Serengeti Command-Line Interface Client

If there is a disk failure in a Hadoop cluster, and the disk does not perform management roles such as NameNode, JobTracker, ResourceManager, HMaster, or ZooKeeper, you can recover by running the Serengeti cluster fix command.

Big Data Extensions uses a large number of inexpensive disk drives for data storage (configured as Just a Bunch of Disks). If several disks fail, the Hadoop data node might shutdown. Big Data Extensions lets you to recover from disk failures.

VMware, Inc. 51

VMware vSphere Big Data Extensions Command-Line Interface Guide

Serengeti supports recovery from swap and data disk failure on all supported Hadoop distributions. Disks are recovered and started in sequence to avoid the temporary loss of multiple nodes at once. A new disk matches the corresponding failed disk’s storage type and placement policies.

The MapR distribution does not support recovery from disk failure by using the cluster fix command.

IMPORTANT Even if you changed the user password on the cluster's nodes, the changed password is not used for the new nodes that are created by the disk recovery operation. If you set the cluster's initial administrator password when you created the cluster, that initial administrator password is used for the new nodes. If you did not set the cluster's initial administrator password when you created the cluster, new random passwords are used for the new nodes.

Procedure

1 Access the Serengeti CLI.

2 Run the cluster fix command.

The nodeGroup parameter is optional.

cluster fix --name cluster_name --disk [--nodeGroup nodegroup_name]

52 VMware, Inc.

Monitoring the Big Data Extensions

Environment 5

You can monitor the status of Serengeti-deployed clusters, including their datastores, networks, and resource pools through the Serengeti Command-Line Interface. You can also view a list of available Hadoop distributions. Monitoring capabilities are also in the vSphere Web Client.

View Available Hadoop Distributions with the Serengeti Command-Line Interface on page 53

n

You can view a list of Hadoop distributions that are available in your Serengeti deployment. When you create clusters, you can use any available Hadoop distribution.

View Provisioned Hadoop and HBase Clusters with the Serengeti Command-Line Interface on

n

page 54

From the Serengeti Command-Line Interface, you can list the provisioned Hadoop and HBase clusters that are in the Serengeti deployment.

View Datastores with the Serengeti Command-Line Interface on page 54

n

From the Serengeti CLI, you can see the datastores that are in the Serengeti deployment.

View Networks with the Serengeti Command-Line Interface on page 54

n

From the Serengeti CLI, you can see the networks that are in the Serengeti deployment.

View Resource Pools with the Serengeti Command-Line Interface on page 55

n

From the Serengeti CLI, you can see the resource pools that are in the Serengeti deployment.

View Available Hadoop Distributions with the Serengeti CommandLine Interface

You can view a list of Hadoop distributions that are available in your Serengeti deployment. When you create clusters, you can use any available Hadoop distribution.

Procedure

1 Access the Serengeti CLI.

2 Run the distro list command.

The available Hadoop distributions are listed, along with their packages.

What to do next

Before you use a distribution, verify that it includes the services that you want to deploy. If services are missing , add the appropriate packages to the distribution.

VMware, Inc.

53

VMware vSphere Big Data Extensions Command-Line Interface Guide

View Provisioned Hadoop and HBase Clusters with the Serengeti Command-Line Interface

From the Serengeti Command-Line Interface, you can list the provisioned Hadoop and HBase clusters that are in the Serengeti deployment.

Procedure

1 Access the Serengeti CLI.

2 Run the cluster list command.

cluster list

This example displays a specific cluster by including the --name parameter.

cluster list --name cluster_name

This example displays detailed information about a specific cluster by including the --name and --

detail parameters.

cluster list --name cluster_name –-detail

View Datastores with the Serengeti Command-Line Interface

From the Serengeti CLI, you can see the datastores that are in the Serengeti deployment.

Procedure

1 Access the Serengeti CLI.

2 Run the datastore list command.

This example displays detailed information by including the --detail parameter.

datastore list --detail

This example displays detailed information about a specific datastore by including the --name and --

detail parameters.

datastore list --name datastore_name --detail

View Networks with the Serengeti Command-Line Interface

From the Serengeti CLI, you can see the networks that are in the Serengeti deployment.

Procedure

1 Access the Serengeti CLI.

2 Run the network list command.

This example displays detailed information by including the --detail parameter.

network list --detail

This example displays detailed information about a specific network by including the --name and --

detail parameters.

network list --name network_name --detail

54 VMware, Inc.

Chapter 5 Monitoring the Big Data Extensions Environment

View Resource Pools with the Serengeti Command-Line Interface

From the Serengeti CLI, you can see the resource pools that are in the Serengeti deployment.

Procedure

1 Access the Serengeti CLI.

2 Run the resourcepool list command.

This example displays detailed information by including the --detail parameter.

resourcepool list --detail

This example displays detailed information about a specific datastore by including the --name and --

detail parameters.

resourcepool list --name resourcepool_name –-detail

VMware, Inc. 55

VMware vSphere Big Data Extensions Command-Line Interface Guide

56 VMware, Inc.

Using Hadoop Clusters from the

Serengeti Command-Line Interface 6

The Serengeti Command-Line Interface lets you perform Hadoop operations. You can run Hive and Pig scripts, HDFS commands, and MapReduce jobs.

The procedures in this section describe how to use the Serengeti Command-Line Interface, which is typically used by administrators. You can also access the client nodes that are provisioned by the Serengeti Management Server to perform standard Hadoop operations such as moving files and running jobs. And if you have other systems integrated with your Hadoop clusters, those systems can directly communicate with the clusters.

Run HDFS Commands with the Serengeti Command-Line Interface on page 57

n

You can run Hadoop Distributed File System (HDFS) commands from the Serengeti Command-Line Interface Client. The HDFS commands let you directly interact with the HDFS as well as other file systems that Hadoop supports.

Run MapReduce Jobs with the Serengeti Command-Line Interface on page 58

n

You can run MapReduce jobs on your Hadoop cluster.

Run Pig and PigLatin Scripts with the Serengeti Command-Line Interface on page 58

n

You can run Pig and PigLatin scripts from the Serengeti Command-Line Interface.

Run Hive and Hive Query Language Scripts with the Serengeti Command-Line Interface on page 59

n

You can run Hive and Hive Query Language (HQL) scripts from the Serengeti Command-Line Interface Client.

Run HDFS Commands with the Serengeti Command-Line Interface

You can run Hadoop Distributed File System (HDFS) commands from the Serengeti Command-Line Interface Client. The HDFS commands let you directly interact with the HDFS as well as other file systems that Hadoop supports.

The File System (FS) shell provides various commands that let you interact with the HDFS, as well as other file systems that Hadoop supports such as Local FS, HFTP FS, S3 FS, and others. You invoke the FS shell using the fs command.

Procedure

1 Access the Serengeti CLI.

2 Run the cluster list command to list the available clusters.

3 Connect to the target cluster.

cluster target –-name cluster_name

VMware, Inc.

57

VMware vSphere Big Data Extensions Command-Line Interface Guide

4 Run HDFS commands.

This example uses the fs put command to move files from /home/serengeti/data to the HDFS path /tmp.

fs put –-from /home/serengeti/data –-to /tmp

This example uses the fs get command to download a file from a specific HDFS to your local filesystem.

fs get --from source_path_and_file --to dest_path_and_file

This example uses the fs ls command to display the contents of the dir1 directory in /user/serengeti.

fs ls /user/serengeti/dir1

This example uses the fs mkdir command to create the dir2 directory in the /user/serengeti directory.

fs mkdir /user/serengeti/dir2

Run MapReduce Jobs with the Serengeti Command-Line Interface

You can run MapReduce jobs on your Hadoop cluster.

Procedure

1 Access the Serengeti CLI, and connect to a Hadoop cluster.

2 Show available clusters.

cluster list

3 Connect to the cluster where you want to run a MapReduce job.

cluster target --name cluster_name

4 Run the mr jar command.

mr jar --jarfile path_and_jar_filename --mainclass class_with_main_method [--args double_quoted_arg_list]

This example runs the MapReduce job located in the hadoop-examples-1.2.1.jar JAR, which is located in the /serengeti/cli/lib directory. Two arguments pass to the MapReduce job: /tmp/input and /tmp/output.

mr jar --jarfile /opt/serengeti/cli/lib/hadoop-examples-1.2.1.jar --mainclass org.apache.hadoop.examples.WordCount --args "/tmp/input /tmp/output"

5 Show the output of the MapReduce job.

fs cat file_to_display_to_stdout

6 Download the output of the MapReduce job from HDFS to the local file system.

fs get --from HDFS_file_path_and_name --to local_file_path_and_name

Run Pig and PigLatin Scripts with the Serengeti Command-Line Interface

You can run Pig and PigLatin scripts from the Serengeti Command-Line Interface.

Pig lets you write PigLatin statements. PigLatin statements are converted by the Pig service into MapReduce jobs, which are executed across your Hadoop cluster.

Prerequisites

Create a PigLatin script to execute against your Hadoop cluster.

58 VMware, Inc.

Chapter 6 Using Hadoop Clusters from the Serengeti Command-Line Interface

Procedure

1 Access the Serengeti CLI.

2 Show available clusters.

cluster list

3 Connect to the cluster where you want to run a Pig script.

cluster target --name cluster_name

4 Run the pig script command to run an existing PigLatin script.

This example runs the PigLatin script data.pig located in the /pig/scripts directory.

pig script --location /pig/scripts/data.pig

What to do next

If the PigLatin script stores its results in a file, you might want to copy that file to your local file system.

Run Hive and Hive Query Language Scripts with the Serengeti Command-Line Interface

You can run Hive and Hive Query Language (HQL) scripts from the Serengeti Command-Line Interface Client.

Hive lets you write HQL statements. HQL statements are converted by the Hive service into MapReduce jobs, which are executed across your Hadoop cluster.

Prerequisites

Create an HQL script to execute against your Hadoop cluster.

Procedure

1 Access the Serengeti CLI.

2 Show available clusters by running the cluster list command.

cluster list

3 Connect to the cluster where you want to run a Hive script.

cluster target --name cluster_name

4 Run the hive script command to run an existing Hive script.

This example runs the Hive script data.hive located in the /hive/scripts directory. hive script --

location /hive/scripts/hive.data

VMware, Inc. 59

VMware vSphere Big Data Extensions Command-Line Interface Guide

60 VMware, Inc.

Cluster Specification Reference 7

This information describes the Serengeti cluster specification file’s attributes and their mapping to Hadoop attributes, and how to convert a Hadoop XML configuration file to a Serengeti configuration file.

Cluster Specification File Requirements on page 61

n

A cluster specification file is a text file with the configuration attributes provided in a JSON-like formatted structure. Cluster specification files must adhere to requirements concerning syntax, quotation mark usage, and comments.

Cluster Definition Requirements on page 62

n

Cluster specification files contain configuration definitions for clusters, such as their roles and node groups. Cluster definitions must adhere to requirements concerning node group roles, cluster roles, and instance numbers.

Annotated Cluster Specification File on page 62

n

The Serengeti cluster specification file defines the nodes, resources, and so on for a cluster. You can use this annotated cluster specification file, and the sample files in /opt/serengeti/samples, as models when you create your clusters.

Cluster Specification Attribute Definitions on page 66

n

Cluster definitions include attributes for the cluster itself and for each of the cluster's node groups.

White Listed and Black Listed Hadoop Attributes on page 68

n

White listed attributes are Apache Hadoop attributes that you can configure from Serengeti with the

cluster config command. The majority of Apache Hadoop attributes are white listed. However, there

are a few black listed Apache Hadoop attributes, which you cannot configure from Serengeti.

Convert Hadoop XML Files to Serengeti JSON Files on page 70

n

If you defined a lot of attributes in your Hadoop configuration files, you can convert that configuration information into the JSON format that Serengeti can use.

Cluster Specification File Requirements

A cluster specification file is a text file with the configuration attributes provided in a JSON-like formatted structure. Cluster specification files must adhere to requirements concerning syntax, quotation mark usage, and comments.

To parse cluster specification files, Serengeti uses the Jackson JSON Processor. For syntax requirements,

n

such as the truncation policy for float types, see the Jackson JSON Processor Wiki.

VMware, Inc.

61

VMware vSphere Big Data Extensions Command-Line Interface Guide

Always enclose digital values in quotation marks. For example:

n

"mapred.tasktracker.reduce.tasks.maximum" : "2"

The quotation marks ensure that integers are correctly interpreted instead of being converted to doubleprecision floating point, which can cause unintended consequences.

Do not include any comments.

n

Cluster Definition Requirements

Cluster specification files contain configuration definitions for clusters, such as their roles and node groups. Cluster definitions must adhere to requirements concerning node group roles, cluster roles, and instance numbers.

A cluster definition has the following requirements:

Node group roles cannot be empty. You can determine the valid role names for your Hadoop

n

distribution by using the distro list command.

The hadoop_namenode and hadoop_jobtracker roles must be configured in a single node group.

n

In Hadoop 2.0 clusters, such as CDH4 or Pivotal HD, the instance number can be greater than 1 to

n

create an HDFS HA or Federation cluster.

Otherwise, the total instance number must be 1.

n

Node group instance numbers must be positive numbers.

n

Annotated Cluster Specification File

The Serengeti cluster specification file defines the nodes, resources, and so on for a cluster. You can use this annotated cluster specification file, and the sample files in /opt/serengeti/samples, as models when you create your clusters.

The following code is a typical cluster specification file. For code annotations, see Table 7-1.

1 { 2 "nodeGroups" : [ 3 { 4 "name": "master", 5 "roles": [ 6 "hadoop_namenode", 7 "hadoop_jobtracker" 8 ], 9 "instanceNum": 1, 10 "instanceType": "LARGE", 11 "cpuNum": 2, 12 "memCapacityMB":4096, 13 "storage": { 14 "type": "SHARED", 15 "sizeGB": 20 16 }, 17 "haFlag":"on", 18 "rpNames": [ 19 "rp1" 20 ] 21 }, 22 { 23 "name": "data", 24 "roles": [

62 VMware, Inc.

25 "hadoop_datanode" 26 ], 27 "instanceNum": 3, 28 "instanceType": "MEDIUM", 29 "cpuNum": 2, 30 "memCapacityMB":2048, 31 "storage": { 32 "type": "LOCAL", 33 "sizeGB": 50, 34 "dsNames4Data": ["DSLOCALSSD"], 35 "dsNames4System": ["DSNDFS"] 36 } 37 "placementPolicies": { 38 "instancePerHost": 1, 39 "groupRacks": { 40 "type": "ROUNDROBIN", 41 "racks": ["rack1", "rack2", "rack3"] 42 } 43 } 44 }, 45 { 46 "name": "compute", 47 "roles": [ 48 "hadoop_tasktracker" 49 ], 50 "instanceNum": 6, 51 "instanceType": "SMALL", 52 "cpuNum": 2, 53 "memCapacityMB":2048, 54 "storage": { 55 "type": "LOCAL", 56 "sizeGB": 10 57 } 58 "placementPolicies": { 59 "instancePerHost": 2, 60 "groupAssociations": [{ 61 "reference": "data", 62 "type": "STRICT" 63 }] 64 } 65 }, 66 { 67 "name": "client", 68 "roles": [ 69 "hadoop_client", 70 "hive", 71 "hive_server", 72 "pig" 73 ], 74 "instanceNum": 1, 75 "instanceType": "SMALL", 76 "memCapacityMB": 2048, 77 "storage": { 78 "type": "LOCAL", 79 "sizeGB": 10,

Chapter 7 Cluster Specification Reference

VMware, Inc. 63

VMware vSphere Big Data Extensions Command-Line Interface Guide

80 "dsNames": [“ds1”, “ds2”] 81 } 82 } 83 ], 84 "configuration": { 85 } 86 }

The cluster definition elements are defined in the table.

Table 7‑1. Example Cluster Specification Annotation

Line(s) Attribute Example Value Description

4 name master Node group name.

5-8 role hadoop_namenode,

9 instanceNum 1 Number of instances in the

10 instanceType LARGE Node group instance type.

hadoop_jobtracker

Node group role.

hadoop_namenode and hadoop_jobtracker are

deployed to the node group's virtual machine.

node group.

Only one virtual machine is created for the group.

You can have multiple

n

instances for hadoop_tasktracker, hadoop_datanode, hadoop_client, pig, and hive.

For HDFS1 clusters, you

n

can have only one instance of hadoop_namenode and hadoop_jobtracker.

For HDFS2 clusters, you

n

can have two hadoop_namenode instances.

With a MapR

n

distribution, you can configure multiple instances of hadoop_jobtracker.

Instance types are predefined virtual machine specifications, which are combinations of the number of CPUs, RAM sizes, and storage size. The predefined numbers can be overridden by the cpuNum, memCapacityMB, and storage attributes in the Serengeti server specification file.

64 VMware, Inc.

Chapter 7 Cluster Specification Reference

Table 7‑1. Example Cluster Specification Annotation (Continued)

Line(s) Attribute Example Value Description

11 cpuNum 2 Number of CPUs per virtual

12 memCapacityMB 4096 RAM size, in MB, per virtual

13-16 storage See lines 14-15 for one

group's storage attributes

14 type SHARED Storage type.

15 sizeGB 20 Storage size.

17 haFlag on HA protection for the node

18-20 rpNames rp1 Resource pools under which

22-36 Node group definition for

the data node

37-44 placementPolicies See code sample Data node group's

45-57 Node group definition for

the compute node

machine.

This attribute overrides the number of vCPUs in the predefined virtual machine specification.

machine.

This attribute overrides the RAM size in the predefined virtual machine specification.

Node group storage requirements.

The node group is deployed using only shared storage.

Each node in the node group is deployed with 20GB available disk space.

group.

The node group is deployed with vSphere HA protection.

the node group virtual machines are deployed. These pools can be an array of values.

See lines 3-21, which define the same attributes for the master node.

In lines 34-35, data disks are placed on dsNames4Data datastores, and system disks are placed on dsNames4System datastores.

placement policy constraints.

You need at least three ESXi hosts because there are three instances and a requirement that each instance be on its own host. This group is provisioned on hosts on rack1, rack2, and rack3 by using a ROUNDROBIN algorithm.

See lines 4-16, which define the same attributes for the master node.

VMware, Inc. 65

VMware vSphere Big Data Extensions Command-Line Interface Guide

Table 7‑1. Example Cluster Specification Annotation (Continued)

Line(s) Attribute Example Value Description

58-65 placementPolicies See code sample Compute node group's

66-82 Node group definition for

the client node

83-86 configuration Empty in the code sample Hadoop configuration

Cluster Specification Attribute Definitions

placement policy constraints.

You need at least three ESXi hosts to meet the instance requirements. The compute node group references a data node group through STRICT typing. The two compute instances use a data instance on the ESXi host. The STRICT association provides better performance.

See previous node group definitions.

customization.

Cluster definitions include attributes for the cluster itself and for each of the cluster's node groups.

Cluster Specification Outer Attributes

Cluster specification outer attributes apply to the cluster as a whole.

Table 7‑2. Cluster Specification Outer Attributes

Attribute Type Mandatory/ Optional Description

nodeGroups object Mandatory One or more group specifications. See Table 7-3.

configuration object Optional Customizable Hadoop configuration key/value pairs.

externalHDFS string Optional Valid only for compute-only clusters. URI of external HDFS.

Cluster Specification Node Group Objects and Attributes

Node group objects and attributes apply to one node group in a cluster.

Table 7‑3. Cluster Specification’s Node Group Objects and Attributes

Mandatory/

Attribute Type

name string Mandatory User defined node group name.

roles list of string Mandatory List of software packages or services to install on the node group’s

Optional Description

virtual machine. Values must match the roles displayed by the distro list command.

66 VMware, Inc.

Chapter 7 Cluster Specification Reference

Table 7‑3. Cluster Specification’s Node Group Objects and Attributes (Continued)

Mandatory/

Attribute Type

instanceNumber integer Mandatory Number of virtual machines in the node group:

instanceType string Optional Size of virtual machines in the node group, expressed as the name

cpuNum integer Optional

memCapacityMB integer Optional RAM size, in MB, per virtual machine.

Storage object Optional Storage settings.

type string Optional Storage type:

sizeGB integer Optional Data storage size. Must be a positive integer.

dsNames list of string Optional Array of datastores the node group can use.

dnNames4Data list of string Optional Array of datastores the data node group can use.

dsNames4System list of string Optional Array of datastores the system can use.

rpNames list of string Optional Array of resource pools the node group can use.

haFlag string Optional By default, NameNode and JobTracker nodes are protected by

placementPolicies object Optional Up to three optional constraints:

Optional Description

Positive integer.

n

Generally, you can have multiple instances for

n

hadoop_tasktracker, hadoop_datanode, hadoop_client, pig, and hive.

For HDFS1 clusters, you can have only one instance of

n

hadoop_namenode and hadoop_jobtracker.

n

For HDFS2 clusters, you can have two hadoop_namenode instances.

With a MapR distribution, you can configure multiple

n

instances of hadoop_jobtracker.

of a predefined virtual machine template. See Table 7-4.

SMALL

n

MEDIUM

n

LARGE

n

EXTRA_LARGE

n

If you specify cpuNum , memCapacityMB , or sizeGB attributes, they override the corresponding value of your selected virtual machine template for the applicable node group.

Number of CPUs per virtual machine. If the haFlag value is FT, the cpuNum value must be 1.

NOTE When using MapR 3.1, you must specify a minimum of 5120 MBs of memory capacity for the zookeeper, worker, and client nodes.

LOCAL. For local storage

n

SHARED. For shared storage.

n

vSphere HA.

on. Protect the node with vSphere HA.

n

ft. Protect the node with vSphere FT.

n

off. Do not use vSphere HA or vSphere FT.

n

instancePerHost

n

groupRacks

n

groupAssociations

n

VMware, Inc. 67

VMware vSphere Big Data Extensions Command-Line Interface Guide

Serengeti Predefined Virtual Machine Sizes

Serengeti provides predefined virtual machine sizes to use for defining the size of virtual machines in a cluster node group.

Table 7‑4. Serengeti Predefined Virtual Machine Sizes

SMALL MEDIUM LARGE EXTRA_LARGE

Number of CPUs per virtual machine

RAM, in GB 3.75 7.5 15 30

Hadoop master data disk size, in GB

Hadoop worker data disk size, in GB

Hadoop client data disk size, in GB

Zookeeper data disk size, in GB

1 2 4 8

25 50 100 200

50 100 200 400

20 40 80 120

White Listed and Black Listed Hadoop Attributes

White listed attributes are Apache Hadoop attributes that you can configure from Serengeti with the

cluster config command. The majority of Apache Hadoop attributes are white listed. However, there are a

few black listed Apache Hadoop attributes, which you cannot configure from Serengeti.

If you use an attribute in the cluster specification file that is neither a white listed nor a black listed attribute, and then run the cluster config command, a warning appears and you must answer yes to continue or no to cancel.

If your cluster includes a NameNode or JobTracker, Serengeti configures the fs.default.name and

dfs.http.address attributes. You can override these attributes by defining them in your cluster

specification.

Table 7‑5. Configuration Attribute White List

File Attributes

core-site.xml All core-default configuration attributes listed on the Apache Hadoop 2.x documentation Web

page. For example, http://hadoop.apache.org/docs/branch_name/core-default.html.

Exclude the attributes defined in the black list.

hdfs-site.xml All hdfs-default configuration attributes listed on the Apache Hadoop 2.x documentation Web

page. For example, http://hadoop.apache.org/docs/branch_name/hdfs-default.html.

Exclude the attributes defined in the black list.

mapred-site.xml All mapred-default configuration attributes listed on the Apache Hadoop 2.x documentation

Web page. For example, http://hadoop.apache.org/docs/branch_name/mapred- default.html.

Exclude the attributes defined in the black list.

68 VMware, Inc.

Table 7‑5. Configuration Attribute White List (Continued)

File Attributes

hadoop-env.sh

log4j.properties

fairscheduler.xml

capacityscheduler.xml

mapred-queueacls.xml

JAVA_HOME

PATH

HADOOP_CLASSPATH

HADOOP_HEAPSIZE

HADOOP_NAMENODE_OPTS

HADOOP_DATANODE_OPTS

HADOOP_SECONDARYNAMENODE_OPTS

HADOOP_JOBTRACKER_OPTS

HADOOP_TASKTRACKER_OPTS

HADOOP_LOG_DIR

hadoop.root.logger

hadoop.security.logger

log4j.appender.DRFA.MaxBackupIndex

log4j.appender.RFA.MaxBackupIndex

log4j.appender.RFA.MaxFileSize

text All fair_scheduler configuration attributes listed on the Apache Hadoop 2.x documentation

Web page that can be used inside the text field. For example, http://hadoop.apache.org/docs/branch_name/fair_scheduler.html.

Exclude the attributes defined in the black list.

All capacity_scheduler configuration attributes listed on the Apache Hadoop 2.x documentation Web page. For example, http://hadoop.apache.org/docs/branch_name/capacity_scheduler.html.

Exclude attributes defined in black list

All mapred-queue-acls configuration attributes listed on the Apache Hadoop 2.x Web page. For example,

http://hadoop.apache.org/docs/branch_name/cluster_setup.html#Configuring+the +Hadoop+Daemons.

Exclude the attributes defined in the black list.

Chapter 7 Cluster Specification Reference

Table 7‑6. Configuration Attribute Black List

File Attributes

core-site.xml

hdfs-site.xml

mapred-site.xml

hadoop-env.sh

VMware, Inc. 69

net.topology.impl

net.topology.nodegroup.aware

dfs.block.replicator.classname

topology.script.file.name

dfs.http.address

dfs.name.dir

dfs.data.dir

mapred.job.tracker

mapred.local.dir

mapred.task.cache.levels

mapred.jobtracker.jobSchedulable

mapred.jobtracker.nodegroup.aware

HADOOP_HOME

HADOOP_COMMON_HOME

HADOOP_MAPRED_HOME

HADOOP_HDFS_HOME

HADOOP_CONF_DIR

HADOOP_PID_DIR

VMware vSphere Big Data Extensions Command-Line Interface Guide

Table 7‑6. Configuration Attribute Black List (Continued)

File Attributes

log4j.properties

fair-scheduler.xml

capacity-scheduler.xml

mapred-queue-acls.xml

None

Convert Hadoop XML Files to Serengeti JSON Files

If you defined a lot of attributes in your Hadoop configuration files, you can convert that configuration information into the JSON format that Serengeti can use.

Procedure

1 Copy the directory $HADOOP_HOME/conf/ from your Hadoop cluster to the Serengeti Management Server.

2 Open a command shell, such as Bash or PuTTY, log in to the Serengeti Management Server, and run the

convert-hadoop-conf.rb Ruby conversion script.

convert-hadoop-conf.rb path_to_hadoop_conf

The converted Hadoop configuration attributes, in JSON format, appear.

3 Open the cluster specification file for editing.

4 Replace the cluster level configuration or group level configuration items with the output that was

generated by the convert-hadoop-conf.rb Ruby conversion script.

What to do next

Access the Serengeti CLI, and use the new specification file.

To apply the new configuration to a cluster, run the cluster config command. Include the --specFile

n

parameter and its value: the new specification file.

To create a cluster with the new configuration, run the cluster create command. Include the --

n

specFile parameter and its value: the new specification file.

70 VMware, Inc.

Serengeti CLI Command Reference 8

This section provides descriptions and syntax requirements for every Serengeti CLI command.

cfg Commands on page 72

n

The cfg {*} commands let you view and specify configuration information for your MapReduce jobs. These commands let you set the configuration properties and their values in the core-site.xml, hdfs-

site.xml, and mapred-site.xml configuration files for jobs started from the command-line interface

using mr commands.

cluster Commands on page 74

n

The cluster {*} commands let you connect to Hadoop and HBase clusters, create and delete clusters, stop and start clusters, and perform cluster management operations.

connect Command on page 80

n

The connect command lets you connect and log in to a remote Serengeti server.

datastore Commands on page 81

n

The datastore {*} commands let you add and delete datastores, and view the list of datastores in a Serengeti deployment.

disconnect Command on page 82

n

The disconnect command lets you disconnect and log out from a remote Serengeti server. After you disconnect from the server, you cannot run any Serengeti commands until you reconnect with the

connect command.

VMware, Inc.

distro list Command on page 82

n

The distro list command lets you view the list of roles in a Hadoop distribution.

fs Commands on page 82

n

The fs {*} FileSystem (FS) shell commands let you manage files on the HDFS and local systems. Before you can run an FS command in a Command-Line Interface session, or after the 30 minute session timeout, you must run the cluster target command

hive script Command on page 88

n

The hive script command lets you run a Hive or Hive Query Language (HQL) script.

mr Commands on page 89

n

The mr {*} commands let you manage MapReduce jobs.

network Commands on page 92

n

The network {*} commands let you manage your networks.

pig script Command on page 94

n

The pig script command lets you run a Pig or PigLatin script.

71

VMware vSphere Big Data Extensions Command-Line Interface Guide

resourcepool Commands on page 94

n

The resourcepool {*} commands let you manage resource pools.

topology Commands on page 95

n

The topology {*} commands let you manage cluster topology.

cfg Commands

The cfg {*} commands let you view and specify configuration information for your MapReduce jobs. These commands let you set the configuration properties and their values in the core-site.xml, hdfs-site.xml, and mapred-site.xml configuration files for jobs started from the command-line interface using mr commands.

cfg fs Command on page 72

n

The cfg fs command lets you specify the node running the Hadoop NameNode service.

cfg info Command on page 73

n

The cfg info command lets you get information about your Hadoop cluster's configuration.

cfg jt Command on page 73

n

The cfg jt command lets you specify which node to use for the Hadoop JobTracker service.

cfg load Command on page 73

n

The cfg load command lets you load (or copy) Hadoop configuration parameters from an exiting configuration file. The resource can be a local Hadoop configuration file such as core-site.xml, hdfs-

site.xml, or mapred-site.xml, or a URL.

cfg props get Command on page 73

n

The cfg props get command lets you view the value of a Hadoop property that you specify.

cfg props list Command on page 73

n

The cfg props list command lets you view the value of all the Hadoop cluster configuration properties.

cfg props set Command on page 73

n

The cfg props set command lets you set the value of the Hadoop property you specify.

cfg fs Command

The cfg fs command lets you specify the node running the Hadoop NameNode service.

The NameNode manages the filesystem namespace. It maintains the filesystem tree and the metadata for all the files and directories in the tree. This information is stored persistently on the local disk in the form of two files: the namespace image and the edit log. The NameNode also knows the DataNodes on which all the blocks for a given file are located.

The cfg fs command sets the variable fs.default.name to the hostname and port number on which you intend to run the NameNode. Typically there is a single master node which runs both the NameNode and HBaseMaster services.

Parameter Mandatory/Optional Description

--namenode hostname-ofnamenode:port_num

Mandatory Specifies the NameNode address,

which can be local or

namenode:port_num

72 VMware, Inc.

Chapter 8 Serengeti CLI Command Reference

cfg info Command

The cfg info command lets you get information about your Hadoop cluster's configuration.

You can use the cfg info command to view information about your Hadoop cluster's configuration.

cfg jt Command

The cfg jt command lets you specify which node to use for the Hadoop JobTracker service.

There are two types of nodes that control the job execution process: a JobTracker and a number of TaskTrackers. The JobTracker coordinates all the jobs run on the system by scheduling tasks to run on TaskTrackers. Tasktrackers run tasks and send progress reports to the JobTracker, which keeps a record of the overall progress of each job. If a task fails, the JobTracker can reschedule it on a different TaskTracker.

The cfg jt command sets the variable mapred.job.tracker to the hostname and port number on which you intend to run the JobTracker.

Parameter Mandatory/Optional Description

--jobtracker hostname-ofjobtracker:port_num

Mandatory Specifies the JobTracker address,

which can be local or

jobtracker:port_num

cfg load Command

The cfg load command lets you load (or copy) Hadoop configuration parameters from an exiting configuration file. The resource can be a local Hadoop configuration file such as core-site.xml, hdfs-

site.xml, or mapred-site.xml, or a URL.

Parameter Mandatory/Optional Description

--location file_path Mandatory Path and file from which to load the

configuration data. For example: "local_dir/hdfs-site.xml". You can also specify the URL of a file.

cfg props get Command

The cfg props get command lets you view the value of a Hadoop property that you specify.

Parameter Mandatory/Optional Description

--key Mandatory The name of the property whose value you want to view. For example:

fs.default.name

cfg props list Command

The cfg props list command lets you view the value of all the Hadoop cluster configuration properties.

You can use the cfg props list command to view the value of all Hadoop cluster configuration properties.

cfg props set Command

The cfg props set command lets you set the value of the Hadoop property you specify.

VMware, Inc. 73

VMware vSphere Big Data Extensions Command-Line Interface Guide

Parameter Mandatory/Optional Description

--property name=value Mandatory Specify the property name and the

cluster Commands

The cluster {*} commands let you connect to Hadoop and HBase clusters, create and delete clusters, stop and start clusters, and perform cluster management operations.

cluster config Command on page 75

n

The cluster config command lets you modify the configuration of an existing Hadoop or HBase cluster, whether the cluster is the configured according to the Serengeti defaults or you have customized the cluster.

cluster create Command on page 75

n

The cluster create command lets you create a Hadoop or HBase cluster.

cluster delete Command on page 76

n

The cluster delete command lets you delete a Hadoop or HBase cluster in Serengeti. When a cluster is deleted, all its virtual machines and resource pools are destroyed.

cluster export Command on page 76

n

The cluster export command lets you export cluster configuration information to a cluster specification file or to display the cluster configuration information to the console.

value to assign.

cluster fix Command on page 77

n

The cluster fix command lets you recover from a failed disk.

cluster list Command on page 77

n

The cluster list command lets you view a list of provisioned clusters in Serengeti. You can see the following information: name, distribution, status, and each node group's information. The node group information consists of the instance count, CPU, memory, type, and size.

cluster resetParam Command on page 77

n

The cluster resetParam command lets you reset a cluster’s scaling parameters and ioShares level to default values. You must specify at least one optional parameter.

cluster resize Command on page 78

n

The cluster resize command lets you change the number of nodes in a node group or scale up/down cluster CPU or RAM. You must specify at least one optional parameter.

cluster setParam Command on page 78

n

The cluster setParam command lets you set scaling parameters and the ioShares priority for a Hadoop cluster in Serengeti. You must specify at least one optional parameter.

cluster start Command on page 79

n

The cluster start command lets you start a cluster in Serengeti.

cluster stop Command on page 79

n

The cluster stop command lets you stop a Hadoop cluster in Serengeti.

cluster target Command on page 80

n

The cluster target command lets you connect to a cluster to run fs, mr, pig, and hive commands with the Serengeti CLI. You must rerun the cluster target command if it has been more than 30 minutes since you ran it in your current Serengeti Command-Line Interface session.

74 VMware, Inc.

Chapter 8 Serengeti CLI Command Reference

cluster upgrade Command on page 80

n

The cluster upgrade command lets you upgrade the components in each Big Data cluster's virtual machines created in a previous version of Big Data Extensions.

cluster config Command

The cluster config command lets you modify the configuration of an existing Hadoop or HBase cluster, whether the cluster is the configured according to the Serengeti defaults or you have customized the cluster.

You can use the cluster config command with the cluster export command to return cluster services and the original Hadoop configuration to normal in the following situations:

A service such as NameNode, JobTracker, DataNode, or TaskTracker goes down.

n

You manually changed the Hadoop configuration of one or more of the cluster's nodes.

n

Run the cluster export command, and then run the cluster config command. Include the new cluster specification file that you just exported.

Table 8‑1.

Parameter Mandatory/Optional Description

--name cluster_name_in_Serengeti

--specFile spec_file_path

--yes

--skipConfigValidation

Mandatory Name of Hadoop cluster to configure.

Optional File name of Hadoop cluster specification

Optional Answer Y to Y/N confirmation. If not specified, manually

type y or n.

Optional Skip cluster configuration validation.

cluster create Command

The cluster create command lets you create a Hadoop or HBase cluster.

If the cluster specification does not include the required nodes, for example a master node, Serengeti creates the cluster according to the default Serengeti-deployed cluster configuration.

When you create a MapReduce cluster with the Serengeti Command-Line Interface, by default you create a MapReduce v1 cluster. To create a MapReduce v2 (YARN) cluster, create a cluster specification file modeled after the /opt/serengeti/samples/default_hadoop_yarn_cluster.json file, and specify the --specFile parameter and your cluster specification file in the cluster create ... command.

Table 8‑2.

Parameter Mandatory/Optional Description

--name cluster_name_in_Serengeti

--type cluster_type

--password

--specFile spec_file_path

--distro Hadoop_distro_name

Mandatory Cluster name

Optional Cluster type:

Hadoop, which is the default

n

HBase

n

Optional Do not use if you use the --

resume parameter

Optional Cluster specification filename

Optional Hadoop distribution for the cluster.

Custom password for all the nodes in the cluster.

Passwords are from 8 to 128 characters, and include only alphanumeric characters ([0-9, a-z, A-Z]) and the following special characters: _ @ # $ % ^ & *

VMware, Inc. 75

VMware vSphere Big Data Extensions Command-Line Interface Guide

Table 8‑2. (Continued)

Parameter Mandatory/Optional Description

--dsNames datastore_names

--networkName

management_network_name

--hdfsNetworkName

hdfs_network_name

--mapredNetworkName

mapred_network_name

--rpNames resource_pool_name

--resume

--topology topology_type

--yes

--skipConfigValidation

Optional Datastore to use to deploy Hadoop cluster in

Mandatory

If you omit any of the optional network parameters, the traffic associated with that parameter is routed on the management network that you specify by the

--networkName parameter.

Optional Network to use for HDFS traffic in Hadoop

Optional Network to use for MapReduce traffic in

Optional Resource pool to use for Hadoop clusters.

Optional Do not use if you use the --

password parameter

Optional Topology type for rack awareness: HVE,

Optional Answer Y to Y/N confirmation. If not specified,

Optional Skip cluster configuration validation.

Serengeti. Multiple datastores can be used, separated by “,”.

By default, all available datastores are used. When you specify the --dsNames parameter, the

cluster can use only those datastores that you provide in this command.

Network to use for management traffic in Hadoop clusters.

clusters.

Hadoop clusters.

Multiple resource pools can be used, separated by “,”.

Recover from a failed deployment process.

RACK_AS_RACK, or HOST_AS_RACK.

manually type y or n.

cluster delete Command

The cluster delete command lets you delete a Hadoop or HBase cluster in Serengeti. When a cluster is deleted, all its virtual machines and resource pools are destroyed.

Table 8‑3.

Parameter Mandatory/Optional Description

--name cluster_name

Mandatory Name of cluster to delete.

cluster export Command

The cluster export command lets you export cluster configuration information to a cluster specification file or to display the cluster configuration information to the console.

Table 8‑4.

Parameter Mandatory/Optional Description

--name cluster_name

--specFile

76 VMware, Inc.

Mandatory Name of cluster to export

Optional File name for exported cluster specification.

If not specified, the output appears in the console.

Chapter 8 Serengeti CLI Command Reference

cluster fix Command

The cluster fix command lets you recover from a failed disk.

IMPORTANT Even if you changed the user password on the cluster's nodes, the changed password is not used for the new nodes that are created by the disk recovery operation. If you set the cluster's initial administrator password when you created the cluster, that initial administrator password is used for the new nodes. If you did not set the cluster's initial administrator password when you created the cluster, new random passwords are used for the new nodes.

Table 8‑5.

Parameter Mandatory/Optional Description

--name cluster_name

--disk

--nodeGroup nodegroup_name

Mandatory Name of cluster that has a failed disk.

Required Recover node disks.

Optional Perform scan and recovery only on the specified node group, not

on all the management nodes in the cluster.

cluster list Command

The cluster list command lets you view a list of provisioned clusters in Serengeti. You can see the following information: name, distribution, status, and each node group's information. The node group information consists of the instance count, CPU, memory, type, and size.

Table 8‑6.

Parameter Mandatory/Optional Description

--name cluster_name_in_Serengeti

--detail

Optional Name of cluster to list.

Optional List cluster details, including name in Serengeti, distribution,

deploy status, each node’s information in different roles.

If you specify this option, Serengeti queries the vCenter Server to get the latest node status.

cluster resetParam Command

The cluster resetParam command lets you reset a cluster’s scaling parameters and ioShares level to default values. You must specify at least one optional parameter.

Table 8‑7.

Parameter Mandatory/Optional Description

--name cluster_name

--all

--elasticityMode

--targetComputeNodeNum

--minComputeNodeNum

Mandatory Name of cluster for which to reset scaling parameters.

Optional Reset all scaling parameters to their defaults.

Optional Reset auto to false.

Optional Reset to -1.

Big Data Extensions manual scaling operations treat a targetComputeNodeNum value of -1 as if it were unspecified upon a change to manual scaling.

Optional Reset to -1.

It appears as "Unset" in the Serengeti CLI displays. Big Data Extensions elastic scaling treats a minComputeNodeNum value of -1 as if it were zero (0).

VMware, Inc. 77

VMware vSphere Big Data Extensions Command-Line Interface Guide

Table 8‑7. (Continued)

Parameter Mandatory/Optional Description

--maxComputeNodeNum

--ioShares

Optional Reset to -1. It appears as "Unset" in the Serengeti CLI displays.

Optional Reset to NORMAL.

cluster resize Command

The cluster resize command lets you change the number of nodes in a node group or scale up/down cluster CPU or RAM. You must specify at least one optional parameter.

If you specify the --instanceNum parameter, you cannot specify either the --cpuNumPerNode parameter or the

--memCapacityMbPerNode parameter.

You can specify the --cpuNumPerNode and the --memCapacityMbPerNode parameters at the same time to scale the CPU and RAM with a single command.

IMPORTANT Even if you changed the user password on the cluster's nodes, the changed password is not used for the new nodes that are created when you scale out a cluster. If you set the cluster's initial administrator password when you created the cluster, that initial administrator password is used for the new nodes. If you did not set the cluster's initial administrator password when you created the cluster, new random passwords are used for the new nodes.

Big Data Extensions elastic scaling treats a maxComputeNodeNum value of -1 as if it were unlimited.

Table 8‑8.

Parameter Mandatory/Optional Description

--name cluster_name_in_Serengeti

--nodeGroup name_of_the_node_group

--instanceNum instance_number

--cpuNumPerNode num_of_vCPUs

--memCapacityMbPerNode size_in_MB

Mandatory Target Hadoop cluster in Serengeti.

Mandatory Target role to scale out in the cluster deployed by

Serengeti.

Optional Target count to which to scale out. Must be greater than

the original count.

Optional Number of vCPUs to use for a node group

Optional Total memory, in MB, to use for the node group

cluster setParam Command

The cluster setParam command lets you set scaling parameters and the ioShares priority for a Hadoop cluster in Serengeti. You must specify at least one optional parameter.

In elastic scaling, Big Data Extensions has two different behaviors for deciding how many active compute nodes to maintain. In both behaviors, Big Data Extensions replaces unresponsive or faulty nodes with live, responsive nodes.

Variable. The number of active, healthy TaskTracker compute nodes is maintained from the configured

n

minimum number of compute nodes to the configured maximum number of compute nodes. The number of active compute nodes varies as resource availability fluctuates.

Fixed. The number of active, healthy TaskTracker compute nodes is maintained at a fixed number when

n

the same value is configured for the minimum and maximum number of compute nodes.

78 VMware, Inc.

Chapter 8 Serengeti CLI Command Reference

Table 8‑9.

Parameter Mandatory/Optional Description

--name cluster_name

--elasticityMode mode

--targetComputeNodeNum

numTargetNodes

--minComputeNodeNum

minNum

--maxComputeNodeNum

maxNum

--ioShares level

Mandatory Name of cluster for which to set elasticity

parameters.

Optional MANUAL or AUTO.

Optional

This parameter is applicable only for the MANUAL scaling mode. If the cluster is in AUTO mode or you are changing it to AUTO mode, this parameter is ignored.

Optional

This parameter is applicable only for the AUTO scaling mode. If the cluster is in MANUAL mode or you are changing it to MANUAL mode, this parameter is ignored.

Optional

This parameter is applicable only for the AUTO scaling mode. If the cluster is in MANUAL mode or you are changing it to MANUAL mode, this parameter is ignored.

Optional Priority access level: LOW, NORMAL, or HIGH.

Number of compute nodes for the specified Hadoop cluster or node group within that cluster.

Must be an integer >= 0.

If zero (0), all the nodes in the specific

n

Hadoop cluster, or if nodeGroup is specified, in the node group, are decommissioned and powered off.

For integers from one to the max number of

n

nodes in the Hadoop cluster, the specified number of nodes remain commissioned and powered on, and the remaining nodes are decommissioned.

For integers > the max number of nodes in

n

the Hadoop cluster, all the nodes in the specified Hadoop cluster or node group are re-commissioned and powered on.

Lower limit of the range of active compute nodes to maintain in the cluster.

Upper limit of the range of active compute nodes to maintain in the cluster.

cluster start Command

The cluster start command lets you start a cluster in Serengeti.

Table 8‑10.

Parameter Mandatory/Optional Description

--name cluster_name

Mandatory Name of cluster to start.

cluster stop Command

The cluster stop command lets you stop a Hadoop cluster in Serengeti.

Table 8‑11.

Parameter Mandatory/Optional Description

--name cluster_name

VMware, Inc. 79

Mandatory Name of cluster to stop.

VMware vSphere Big Data Extensions Command-Line Interface Guide

cluster target Command

The cluster target command lets you connect to a cluster to run fs, mr, pig, and hive commands with the Serengeti CLI. You must rerun the cluster target command if it has been more than 30 minutes since you ran it in your current Serengeti Command-Line Interface session.

Table 8‑12.

Parameter Mandatory/Optional Description

--name

cluster_name

--info

Optional Name of the cluster to which to connect. If unspecified, the first cluster

listed by the cluster list command is connected. The --name and --info parameters are mutually exclusive. You can use

either parameter, but not both.

Optional Show targeted cluster information, such as the HDFS URL, Job Tracker

URL and Hive server URL. The --info and --name parameters are mutually exclusive. You can use

either parameter, but not both.

cluster upgrade Command

The cluster upgrade command lets you upgrade the components in each Big Data cluster's virtual machines created in a previous version of Big Data Extensions.

To enable the Serengeti Management Server to manage clusters created in a previous version of Big Data Extensions, you must upgrade the components in each cluster's virtual machines. The Serengeti Management Server uses these components to control the cluster nodes.

You can identify clusters you need to upgrade using the cluster list command. When you run the cluster

list command the message "Earlier" displays where the cluster version normally appears. See “cluster list

Command,” on page 77.

Table 8‑13.

Parameter Mandatory/Optional Description

--name cluster_name_in_Serengeti

connect Command

The connect command lets you connect and log in to a remote Serengeti server.

The connect command reads the user name and password in interactive mode. You must run the connect command every time you begin a Serengeti Command-Line Interface session, and again after the 30 minute session timeout. If you do not run this command, you cannot run any other commands.

Table 8‑14.

Parameter Mandatory/Optional Description

--host Mandatory Serengeti Web service URL, formatted as

Mandatory Name of cluster to upgrade.

serengeti_management_server_ip_or_host:port. By default, the Serengeti web service is started at port 8443.

80 VMware, Inc.

datastore Commands

The datastore {*} commands let you add and delete datastores, and view the list of datastores in a Serengeti deployment.

datastore add Command on page 81

n

The datastore add command lets you add a datastore to Serengeti.

datastore delete Command on page 81

n

The datastore delete command lets you delete a datastore from Serengeti.

datastore list Command on page 81

n

The datastore list command lets you view a list of datastores in Serengeti. If you do not specify a datastore name, all datastores are listed.

datastore add Command

The datastore add command lets you add a datastore to Serengeti.

Table 8‑15.

Parameter Mandatory/Optional Description

--name datastore_name_in_Serengeti Mandatory Datastore name in Serengeti.

--spec datastore_name_in_vCenter_Server Mandatory Datastore name in vSphere. You can use a wildcard to

--type {LOCAL|SHARED}

Chapter 8 Serengeti CLI Command Reference

specify multiple vmfs stores. Supported wildcards are * and ?.

Optional (Default=SHARED) Type of datastore: LOCAL or

SHARED.

datastore delete Command

The datastore delete command lets you delete a datastore from Serengeti.

Table 8‑16.

Parameter Mandatory/Optional Description

--name datastore_name_in_Serengeti Mandatory Name of datastore to delete.

datastore list Command

The datastore list command lets you view a list of datastores in Serengeti. If you do not specify a datastore name, all datastores are listed.

Table 8‑17.

Parameter Mandatory/Optional Description

--name Name_of_datastore_name_in_Serengeti Optional Name of datastore to list.

--detail Optional List the datastore details, including the datastore

path in vSphere.

VMware, Inc. 81

VMware vSphere Big Data Extensions Command-Line Interface Guide

disconnect Command

The disconnect command lets you disconnect and log out from a remote Serengeti server. After you disconnect from the server, you cannot run any Serengeti commands until you reconnect with the connect command.

There are no command parameters.

distro list Command

The distro list command lets you view the list of roles in a Hadoop distribution.

Table 8‑18.

Parameter Mandatory/Optional Description

--name distro_name

Optional Name of distribution to show.

fs Commands

The fs {*} FileSystem (FS) shell commands let you manage files on the HDFS and local systems. Before you can run an FS command in a Command-Line Interface session, or after the 30 minute session timeout, you must run the cluster target command

fs cat Command on page 83

n

The fs cat command lets you copy source paths to stdout.

fs chgrp Command on page 84

n

The fs chgrp command lets you change group associations of one or more files.

fs chmod Command on page 84

n

The fs chmod command lets you change permissions of one or more files.

fs chown Command on page 84

n

The fs chown command lets you change the owner of one or more files.

fs copyFromLocal Command on page 84

n

The fs copyFromLocal command lets you copy one or more source files from the local file system to the destination file system. The result of this command is the same as the put command.

fs copyToLocal Command on page 85

n

The fs copyToLocal command lets you copy one or more files to the local file system. The result of this command is the same as the get command.

fs copyMergeToLocal Command on page 85

n

The fs copyMergeToLocal command lets you concatenate one or more HDFS files to a local file.

fs count Command on page 85

n

The fs count command lets you count the number of directories, files, bytes, quota, and remaining quota.

fs cp Command on page 85

n

The fs cp command lets you copy one or more files from source to destination. This command allows multiple sources, in which case the destination must be a directory.

fs du Command on page 86

n

The fs du command lets you display the size of files and directories that are in the given directory, or if just a file is specified, the file size.

82 VMware, Inc.

Chapter 8 Serengeti CLI Command Reference

fs expunge Command on page 86

n

The fs expunge command lets you empty the HDFS trash bin. There are no command parameters.

fs get Command on page 86

n

The fs get command lets you copy one or more HDFS files to the local file system.

fs ls Command on page 86

n

The fs lscommand lets you view a list of a directory’s files.

fs mkdir Command on page 86

n

The fs mkdir command lets you create a directory.

fs moveFromLocal Command on page 87

n

The fs moveFromLocal command copies files similarly to the put command, except that the local source file is deleted after it is copied.

fs mv Command on page 87

n

The fs mv command lets you move one or more local source files to an HDFS destination.

fs put Command on page 87

n

The fs put command lets you copy one or more local file system sources to an HDFS.

fs rm Command on page 87

n

The fs rm command lets you remove files from the HDFS.

fs setrep Command on page 88

n

The fs setrep command lets you change a file’s replication factor.

fs tail Command on page 88

n

The fs tail command lets you display a file’s last kilobyte of content to stdout.

fs text Command on page 88

n

The fs text command lets you output a source file in text format.

fs touchz Command on page 88

n

The fs touchz command lets you create a zero length file.

fs cat Command

The fs cat command lets you copy source paths to stdout.

Table 8‑19.

Parameter Mandatory/Optional Description

file_name Mandatory File to be showed in the console.

Multiple files must use quotes. For example: "/path/file1 /path/file2".

VMware, Inc. 83

VMware vSphere Big Data Extensions Command-Line Interface Guide

fs chgrp Command

The fs chgrp command lets you change group associations of one or more files.

Table 8‑20.

Parameter Mandatory/Optional Description

--group group_name

--recursive {true|false}

file_name Mandatory Name of file to change. Multiple files must use quotes. For example:

Mandatory Group name of the file.

Optional Perform the operation recursively.

fs chmod Command

The fs chmod command lets you change permissions of one or more files.

Table 8‑21.

Parameter Mandatory/Optional Description

--mode

--recursive {true|

false}

file_name Mandatory Name of file to change. Multiple files must use quotes. For example:

Mandatory File permission mode. For example: 755.

Optional Perform the operation recursively.

"/path/file1 /path/file2".

"/path/file1 /path/file2"/path/file1 /path/".

fs chown Command

The fs chown command lets you change the owner of one or more files.

Table 8‑22.

Parameter Mandatory/Optional Description

--owner permission_mode

--recursive {true|false}

file_name Mandatory Name of file to change. Multiple files must use quotes. For example:

Mandatory Name of file owner.

Optional Change the owner recursively through the directory structure.

"/path/file1 /path/file2".

fs copyFromLocal Command

The fs copyFromLocal command lets you copy one or more source files from the local file system to the destination file system. The result of this command is the same as the put command.

Table 8‑23.

Parameter Mandatory/Optional Description

--from local_file_path Mandatory Local file path. Multiple files must use quotes. For example:

"/path/file1 /path/file2".

--to HDFS_file_path Mandatory

File path in local. If the from parameter value lists multiple files, the to parameter value is the directory name.

84 VMware, Inc.

Chapter 8 Serengeti CLI Command Reference

fs copyToLocal Command

The fs copyToLocal command lets you copy one or more files to the local file system. The result of this command is the same as the get command.

Table 8‑24.

Parameter Mandatory/Optional Description

--from HDFS_file_path Mandatory File path in HDFS. Multiple files must use quotes. For example:

"/path/file1 /path/file".

--to local_file_path Mandatory

File path in local. If the from parameter value lists multiple files, the to parameter value is the directory name.

fs copyMergeToLocal Command

The fs copyMergeToLocal command lets you concatenate one or more HDFS files to a local file.

Table 8‑25.

Parameter Mandatory/Optional Description

--from HDFS_file_path Mandatory File path in HDFS. Multiple files must use quotes. For example:

"/path/file1 /path/file2".

--to local_file_path Mandatory File path in local.

--endline {true|false}

Optional Add end of line (EOL) character.

fs count Command

The fs count command lets you count the number of directories, files, bytes, quota, and remaining quota.

Table 8‑26.

Parameter Mandatory/Optional Description

--path HDFS_path Mandatory Path to be counted.

--quota {true|false}

Optional Include quota information.

fs cp Command

The fs cp command lets you copy one or more files from source to destination. This command allows multiple sources, in which case the destination must be a directory.

Table 8‑27.

Parameter Mandatory/Optional Description

--from HDFS_source_file_path Mandatory File path in local. Multiple files must use quotes. For example:

"/path/file1 /path/file2".

--to HDFS_destination_file_path Mandatory

File path in local. If the from parameter value lists multiple files, the to parameter value is the directory name.

VMware, Inc. 85

VMware vSphere Big Data Extensions Command-Line Interface Guide

fs du Command

The fs du command lets you display the size of files and directories that are in the given directory, or if just a file is specified, the file size.

Table 8‑28.

Parameter Mandatory/Optional Description

file_name Mandatory File to display. Multiple files must use quotes. For example:

"/path/file1 /path/file2".

fs expunge Command

The fs expunge command lets you empty the HDFS trash bin. There are no command parameters.

fs get Command

The fs get command lets you copy one or more HDFS files to the local file system.

Table 8‑29.

Parameter Mandatory/Optional Description

--from HDFS_file_path Mandatory File path in HDFS. Multiple files must use quotes. For example:

"/path/file1 /path/file2".

--to local_file_path Mandatory

File path in local. If the from parameter value lists multiple files, the to parameter value is the directory name.

fs ls Command

The fs lscommand lets you view a list of a directory’s files.

Table 8‑30.

Parameter Mandatory/Optional Description

path_name Mandatory Path for which to view the list of files. Multiple paths must use

quotes. For example: "/path/file1 /path/file2".

--recursive

{true|false}

Optional Perform the operation recursively.

fs mkdir Command

The fs mkdir command lets you create a directory.

Table 8‑31.

Parameter Mandatory/Optional Description

dir_name Mandatory Name of directory to create.

86 VMware, Inc.

Chapter 8 Serengeti CLI Command Reference

fs moveFromLocal Command

The fs moveFromLocal command copies files similarly to the put command, except that the local source file is deleted after it is copied.

Table 8‑32.

Parameter Mandatory/Optional Description

--from local_file_path Mandatory File path in local. Multiple files must use quotes. For example:

"/path/file1 /path/file2".

--to HDFS_file_path Mandatory

File path in HDFS. If the from parameter value lists multiple files, the to parameter value is the directory name.

fs mv Command

The fs mv command lets you move one or more local source files to an HDFS destination.

Table 8‑33.

Parameter Mandatory/Optional Description

--from source_file_path Mandatory Local source path and file. Multiple files must use quotes. For example:

"/path/file1 /path/file2".

--to HDFS_file_path Mandatory

HDFS destination path and file. If the from parameter value lists multiple files, the to parameter value is the directory name.

fs put Command

The fs put command lets you copy one or more local file system sources to an HDFS.

Table 8‑34.

Parameter Mandatory/Optional Description

--from local_file_path Mandatory Local source path and file. Multiple files must use quotes. For example:

"/path/file1 /path/file2".

--to HDFS_file_path Mandatory

HDFS destination path and file. If the from parameter value lists multiple files, the to parameter value is the directory name.

fs rm Command

The fs rm command lets you remove files from the HDFS.

Table 8‑35.

Parameter Mandatory/Optional Description

file_path Mandatory File to remove.

--recursive {true|false}

--skipTrash {true|false}

Optional Perform the operation recursively.

Optional Bypass trash.

VMware, Inc. 87

VMware vSphere Big Data Extensions Command-Line Interface Guide

fs setrep Command

The fs setrep command lets you change a file’s replication factor.

Table 8‑36.

Parameter Mandatory/Optional Description

--path file_path Mandatory Path and file for which to change the replication factor.

--replica replica_number Mandatory Number of replicas.

--recursive {true|false}

--waiting {true|false}

Optional Perform the operation recursively.

Optional Wait for the replica number to equal the specified number.

fs tail Command

The fs tail command lets you display a file’s last kilobyte of content to stdout.

Table 8‑37.

Parameter Mandatory/Optional Description

file_path Mandatory File path to display.

--file {true|false}

Optional Show content while the file grows.

fs text Command

The fs text command lets you output a source file in text format.

Table 8‑38.

Parameter Mandatory/Optional Description

file_path Mandatory File path to display.

fs touchz Command

The fs touchz command lets you create a zero length file.

Table 8‑39.

Parameter Mandatory/Optional Description

file_path Mandatory Name of file to create.

hive script Command

The hive script command lets you run a Hive or Hive Query Language (HQL) script.

Before you can run the hive script command in a Command-Line Interface session, or after the 30 minute session timeout, you must run the cluster target command.

Table 8‑40.

Parameter Mandatory/Optional Description

--location script_path Mandatory Hive or HQL script file name.

88 VMware, Inc.

mr Commands

The mr {*} commands let you manage MapReduce jobs.

Before you can run an mr {*} command in a Command-Line Interface session, or after the 30 minute session timeout, you must run the cluster target command

mr jar Command on page 89

n

The mr jar command lets you run a MapReduce job located inside the provided jar.

mr job counter Command on page 90

n

The mr job counter command lets you print the counter value of the MapReduce job.

mr job events Command on page 90

n

The mr job events command lets you print the events' details received by JobTracker for the given range.

mr job history Command on page 90

n

The mr job history command lets you print job details for failed and killed jobs.

mr job kill Command on page 90

n

The mr job kill command lets you kill a MapReduce job.

Chapter 8 Serengeti CLI Command Reference

mr job list Command on page 90

n

The mr job list command lets you view the list of MapReduce jobs.

mr job set priority Command on page 91

n

The mr job set priority command lets you change a MapReduce job’s priority.

mr job status Command on page 91

n

The mr job status command lets you query a MapReduce job's status.

mr job submit Command on page 91

n

The mr job submit command lets you submit a MapReduce job that is defined in a MapReduce job file.

mr task fail Command on page 92

n

The mr task fail command lets you fail the MapReduce task.

mr task kill Command on page 92

n

The mr task kill command lets you kill a MapReduce task.

mr jar Command

The mr jar command lets you run a MapReduce job located inside the provided jar.

Table 8‑41.

Parameter Mandatory/Optional Description

--jarfile jar_file_path Mandatory JAR file path.

--mainclass main_class_name Mandatory

--args arg Optional

Class that contains the main() method.

Arguments to send to the main class. To send multiple arguments, double quote them.

VMware, Inc. 89

VMware vSphere Big Data Extensions Command-Line Interface Guide

mr job counter Command

The mr job counter command lets you print the counter value of the MapReduce job.

Table 8‑42.

Parameter Mandatory/Optional Description

--jobid job_id Mandatory MR job id.

--groupname group_name Mandatory Counter’s group name.

--countername counter_name Mandatory Counter’s name.

mr job events Command

The mr job events command lets you print the events' details received by JobTracker for the given range.

Table 8‑43.

Parameter Mandatory/Optional Description

--jobid job_id Mandatory MapReduce job id.

--from from-event-# Mandatory Event number of the first event to print.

--number #-of-events Mandatory Total number of events to print.

mr job history Command

The mr job history command lets you print job details for failed and killed jobs.

Table 8‑44.

Parameter Mandatory/Optional Description

job_history_directory Mandatory Directory into which to place the job history files.

--all {true|false}

Optional Print all jobs information.

mr job kill Command

The mr job kill command lets you kill a MapReduce job.

Table 8‑45.

Parameter Mandatory/Optional Description

--jobid job_id Mandatory Job id of job to kill.

mr job list Command

The mr job list command lets you view the list of MapReduce jobs.

Table 8‑46.

Parameter Mandatory/Optional Description

--all {true|false}

Optional List all jobs.

90 VMware, Inc.

Chapter 8 Serengeti CLI Command Reference

mr job set priority Command

The mr job set priority command lets you change a MapReduce job’s priority.

Table 8‑47.

Parameter Mandatory/Optional Description

--jobid jobid Mandatory Job id of job to change.

--priority {VERY_HIGH|HIGH|NORMAL|LOW|VERY_LOW}

Mandatory Job priority.

mr job status Command

The mr job status command lets you query a MapReduce job's status.

Table 8‑48.

Parameter Mandatory/Optional Description

--jobid jobid Mandatory Job id of job to query.

mr job submit Command

The mr job submit command lets you submit a MapReduce job that is defined in a MapReduce job file.

Table 8‑49.

Parameter Mandatory/Optional Description

--jobfile jobfile Mandatory File that defines the MapReduce job.

Sample MapReduce Job File

A MapReduce job file is a standard Hadoop configuration file.

<configuration> <property> <name>fs.default.name</name> <value>hdfs://localhost:9000</value> </property> <property> <name>mapred.job.tracker</name> <value>localhost:9001</value> </property>

<property> <name>mapred.jar</name> <value>/home/hadoop/hadoop-1.0.1/hadoop-examples-1.0.1.jar</value> </property> <property> <name>mapred.input.dir</name> <value>/user/hadoop/input</value> </property> <property> <name>mapred.output.dir</name> <value>/user/hadoop/output</value> </property> <property>

VMware, Inc. 91

VMware vSphere Big Data Extensions Command-Line Interface Guide

<name>mapred.job.name</name> <value>wordcount</value> </property> <property> <name>mapreduce.map.class</name> <value>org.apache.hadoop.examples.WordCount.TokenizerMapper</value> </property> <property> <name>mapreduce.reduce.class</name> <value>org.apache.hadoop.examples.WordCount.IntSumReducer</value> </property> </configuration>

mr task fail Command

The mr task fail command lets you fail the MapReduce task.

Table 8‑50.

Parameter Mandatory/Optional Description

--taskid taskid Mandatory Task id of task to fail.

mr task kill Command

The mr task kill command lets you kill a MapReduce task.

Table 8‑51.

Parameter Mandatory/Optional Description

--taskid taskid Mandatory Task id of task to kill.

network Commands

The network {*} commands let you manage your networks.

network add Command on page 93

n

The network add command lets you add a network to Serengeti so that The network's IP addresses are available to clusters that you create.

network delete Command on page 93

n

The network delete command lets you delete a network from Serengeti. Deleting an unused network frees the network's IP addresses for use by other services.

network list Command on page 93

n

The network list command lets you view a list of available networks in Serengeti. The name, port group in vSphere, IP address assignment type, assigned IP address, and so on appear.

network modify Command on page 94

n

The network modify command lets you reconfigure a Serengeti static IP network by adding IP address segments to it. You might need to add IP address segments so that there is enough capacity for a cluster that you want to create.

92 VMware, Inc.

Chapter 8 Serengeti CLI Command Reference

network add Command

The network add command lets you add a network to Serengeti so that The network's IP addresses are available to clusters that you create.

NOTE If your network uses static IP addresses, be sure that the addresses are not occupied before you add the network.

This example adds a network with statically assigned IP addresses.

network add --name ipNetwork --ip 192.168.1.1-100,192.168.1.120-180 --portGroup pg1

--dns 202.112.0.1 --gateway 192.168.1.255 --mask 255.255.255.1

This example adds a network with DHCP assigned IP addresses.

network add --name dhcpNetwork --dhcp --portGroup pg1

Specify either the --dhcp parameter for dynamic addresses or the combination of parameters that are required for static addresses, but not parameters for both dynamic and static addresses.

Table 8‑52.

Parameter Mandatory/Optional Description

--name network_name_in_Serengeti

--portGroup port_group_name_in_vSphere

--dhcp

--ip IP_range

--dns dns_server_ip_addr

--secondDNS dns_server_ip_addr

--gateway gateway_IP_addr

--mask network_IP_addr_mask

Mandatory Name of network resource to add.

Mandatory Name of port group in vSphere to

add.

Mandatory for dynamic addresses. Do not use for static addresses.

Mandatory for static addresses. Do not use for dynamic addresses.

Assign dynamic DHCP IP addresses.

Assign static IP addresses.

Express the IP_range in the format xx.xx.xx.xx-xx[,xx]* .

Express IP addresses in the format xx.xx.xx.xx.

network delete Command

The network delete command lets you delete a network from Serengeti. Deleting an unused network frees the network's IP addresses for use by other services.

Table 8‑53.

Parameter Mandatory/Optional Description

--name network_name_in_Serengeti

Mandatory Delete the specified network in Serengeti.

network list Command

The network list command lets you view a list of available networks in Serengeti. The name, port group in vSphere, IP address assignment type, assigned IP address, and so on appear.

Table 8‑54.

Parameter Mandatory/Optional Description

--name network_name_in_Serengeti

--detail

VMware, Inc. 93

Optional Name of network to display.

Optional List network details.

VMware vSphere Big Data Extensions Command-Line Interface Guide

network modify Command

The network modify command lets you reconfigure a Serengeti static IP network by adding IP address segments to it. You might need to add IP address segments so that there is enough capacity for a cluster that you want to create.

NOTE If your network uses static IP addresses, be sure that the addresses are not occupied before you add the network.

Table 8‑55.

Parameter Mandatory/Optional Description

--name network_name_in_Serengeti

--addIP IP_range

Mandatory Modify the specified static IP network

Mandatory IP address segments, in the format

pig script Command

The pig script command lets you run a Pig or PigLatin script.

in Serengeti.

xx.xx.xx.xx-xx[,xx]*.

Before you can run the pig script command in a Command-Line Interface session, or after the 30 minute session timeout, you must run the cluster target command.

Table 8‑56.

Parameter Mandatory/Optional Description

--location script_path

Mandatory Name of the script to execute.

resourcepool Commands

The resourcepool {*} commands let you manage resource pools.

resourcepool add Command on page 94

n

The resourcepool add command lets you add a vSphere resource pool to Serengeti.

resourcepool delete Command on page 95

n

The resourcepool delete command lets you remove a resource pool from Serengeti.

resourcepool list Command on page 95

n

The resourcepool list command lets you view a list of Serengeti resource pools. If you do not specify a name, all Serengeti resource pools are listed.

resourcepool add Command

The resourcepool add command lets you add a vSphere resource pool to Serengeti.

When you add a resource pool in Serengeti, it represents the actual vSphere resource pool as recognized by vCenter Server. This symbolic representation enables you to use the Serengeti resource pool name, instead of the full path of the resource pool in vCenter Server, in cluster specification files.

94 VMware, Inc.

Chapter 8 Serengeti CLI Command Reference

Table 8‑57.

Parameter Mandatory/Optional Description

--name resource_pool_name_in_Serengeti

--vccluster vSphere_cluster_of_the_resource_pool

--vcrp vSphere_resource_pool_name

Mandatory Name of resource pool to add.

Optional Name of vSphere cluster that contains the

resource pool.

Mandatory vSphere resource pool.

resourcepool delete Command

The resourcepool delete command lets you remove a resource pool from Serengeti.

Table 8‑58.

Parameter Mandatory/Optional Description

--name resource_pool_name_in_Serengeti

Mandatory Resource pool to remove.

resourcepool list Command

The resourcepool list command lets you view a list of Serengeti resource pools. If you do not specify a name, all Serengeti resource pools are listed.

Table 8‑59.

Parameter Mandatory/Optional Description

--name resource_pool_name_in_Serengeti Optional Name and path of resource pool to list.

--detail

topology Commands

The topology {*} commands let you manage cluster topology.

topology list Command on page 95

n

The topology list command lets you list the RACK-HOSTS mapping topology stored in Serengeti.

topology upload Command on page 95

n

The topology upload command lets you upload a rack-hosts mapping topology file to Serengeti. The uploaded file overwrites any previous file.

topology list Command

The topology list command lets you list the RACK-HOSTS mapping topology stored in Serengeti.

There are no command parameters.

topology upload Command

Optional Include resource pool details.

The topology upload command lets you upload a rack-hosts mapping topology file to Serengeti. The uploaded file overwrites any previous file.

The file format for every line is: rackname: hostname1, hostname2…

VMware, Inc. 95

VMware vSphere Big Data Extensions Command-Line Interface Guide

Table 8‑60.

Parameter Mandatory/Optional Description

--fileName topology_file_name

--yes

Mandatory Name of topology file.

Optional Answer Y to Y/N confirmation. If not specified, manually type y

or n.

96 VMware, Inc.

Index

A

accessing, Command-Line Interface 7 adding

datastores 10, 81 networks 11 resource pools 10 topology 33

B

balancing workloads 33 basic cluster 29 basic Hadoop clusters 17 black listed Hadoop attributes 68

C

cfg commands 72 cfg fs command 72 cfg info command 73 cfg jt command 73 cfg load command 73 cfg props get 73 cfg props list 73 cfg props set command 73 CLI command reference 71 client nodes for Hadoop 17 Cloudera distribution

administrative commands with the Serengeti

CLI 7

DNS and FQDN for cluster traffic 19, 21–25,

27, 33, 34

cluster create command 18, 75 cluster delete command 51, 76 cluster resize command 42, 43, 78 cluster setParam command 47, 48, 78 cluster target command 58, 80 cluster commands 74 cluster config command 43, 51, 75 cluster export command 43, 76 cluster fix command 51, 77 cluster list command 54, 77 cluster names 18 cluster resetParam command 48, 77 cluster specification files

annotated example 62 cluster definition requirements 62

compute-only cluster 27 data-compute separated clusters 24 defining attributes 66 file requirements 61 Hadoop distribution JARs 43 node group overrides 51 node placement 25 nodes 23 placement policies 32 reconfiguring clusters 43 reference 61 resource pool symbolic link 10 topology 32

topology constraints 34 cluster start command 42, 79 cluster stop command 42, 79 cluster upgrade command 80 clusters

assigning networks 21

assigning resources 20

attributes in definitions 66

basic Hadoop 17

compute-only 17, 27

configuring 23, 75

configuring scaling 48

connecting to 80

creating, See creating clusters

custom administrator passwords 19

customized 17

data-compute separated 17, 24, 25

default Hadoop configuration 18

default HBase configuration 36

defining nodes 66

definition requirements in cluster specification

files 62 definitions, exporting 76 deleting 51, 76 deploying under different resource pools 10 disabling elastic scaling 48 disabling manual scaling 47 elastic scaling 77 enabling elastic scaling 47 enabling manual scaling 48 failover 51

VMware, Inc.

97

VMware vSphere Big Data Extensions Command-Line Interface Guide

Hadoop default 18 HBase 17, 37 managing 41 manual scaling 48, 77 naming 18 node administrator passwords 19 node group roles 22 reconfiguring 43, 51 scaling 78 scaling out 42 scaling parameters 48 starting 42, 79 stopping 42, 79 topology 31, 33, 34 using 57

viewing provisioned 54, 77 Command-Line Interface, accessing 7 compute capacity, scaling 43

compute-only clusters 27 configuration files, converting Hadoop XML to

Serengeti JSON 43

configuring

clusters 23, 75

scaling 48 connect command 80 connecting

clusters 80

Serengeti service 7

to Serengeti servers 80 convert-hadoop-conf.rb conversion tool 43, 70 converting Hadoop XML to Serengeti JSON 43 creating, node administrator passwords 19

creating clusters

compute-only 27

custom administrator password 19

customized 22

data-compute separated 24, 25, 34

default HBase 36

default Hadoop 18

MapReduce v2 21

placement policies 34

placement policy constraints 25

specifying master, worker, and client

nodes 23 topology-aware 33, 34 vSphere HA-protected 37 with assigned networks 21 with assigned resources 20 with available distributions 19

customized clusters, creating 22

D

data-compute separated clusters 17, 24, 25

datastore add command 10, 81 datastore commands 81 datastore delete command 11, 81 datastore list command 11, 54, 81 datastores

adding 10, 81 deleting 81 removing 11

viewing 54, 81 defining, node attributes 66 deleting

clusters 51, 76

datastores 81

See also removing disabling, elastic scaling 48

disconnect command 82 disconnecting, from Serengeti servers 82

disk failure, recovering from 51 distributions, See Hadoop distributions distro list command 53, 82

E

elastic scaling

cluster configuration 77

disabling 48

enabling 47 enabling

elastic scaling 47

manual scaling 48 exporting, cluster definitions 76

F

failed disk, recovering 77 federation 19 fixed elastic scaling 50 fs cat command 83 fs chgrp command 84 fs chmod command 84 fs chown command 84 fs commands 82 fs copyFromLocal command 84 fs copyMergeToLocal command 85 fs copyToLocal command 85 fs count command 85 fs cp command 85 fs du command 86 fs expunge command 86 fs get command 57, 86 fs ls command 57, 86 fs mkdir command 57, 86 fs moveFromLocal command 87 fs mv command 87 fs put command 57, 87

98 VMware, Inc.

Index

fs rm command 87 fs setrep command 88 fs tail command 88 fs text command 88 fs touchz command 88

H

Hadoop clusters

default configuration 18 See also clusters

Hadoop configuration, converting XML to

JSON 70 Hadoop Virtualization Extensions (HVE) 31, 33 Hadoop attributes

black listed 68 white listed 68

Hadoop distributions

configuration files 68 JAR files 43 viewing available 53 viewing list of 82

HBase clusters

configuring 37 creating default 36 default configuration 36 See also clusters

HDFS, avoiding node role conflicts 27 HDFS commands, running 57 Hive scripts, running 59 Hive Query Language 59 hive script command 59, 88 HOST_AS_RACK 31

I

I/O shares 77, 78 IP address segments 12 IP addresses 12

J

Java Runtime Environment (JRE) 7

L

log4j.properties file 43

M

managing

clusters 41 vSphere resources 9

manual scaling

cluster configuration 77 disabling 47 enabling 48

MapR distribution, administrative commands with

the Serengeti CLI 7

MapReduce clusters, creating 21 MapReduce jobs

HBase clusters 37 running 58

See also mr commands master nodes for Hadoop 17 memory, scaling 43

monitoring, Big Data Extensions environment 53 mr commands 89 mr jar command 58, 89 mr job counter command 90 mr job events command 90 mr job history command 90 mr job kill command 90 mr job list command 90 mr job set priority command 91 mr job status command 91 mr job submit command 91 mr task fail command 92 mr task kill command 92

N

network add command 11, 93 network commands 92 network delete command 12, 93 network list command 12, 54, 93 network modify command 94 networks

adding 11

adding IP addresses 12

assigning to clusters 21

removing 12

viewing status 54 node groups

associations 32

in cluster definitions 62

reconfiguring 51

roles, avoiding conflict 27

roles, changing 22 nodes

configuring for elastic scaling 47

configuring in cluster specification files 23

defining attributes 66

distributing 25

scaling out a cluster 42

P

passwords for cluster nodes 19 pig script command 58, 94 Pig scripts, running 58

VMware, Inc. 99

VMware vSphere Big Data Extensions Command-Line Interface Guide

Pivotal distribution

administrative commands with the Serengeti

CLI 7

DNS and FQDN for cluster traffic 19, 21–25,

27, 33, 34

placement policies 25, 32 port groups, See networks predefined virtual machine sizes 66

R

rack topologies 33 RACK_AS_RACK 31 reconfiguring

networks 12

node groups 51 recovering from disk failure 51, 77 removing

datastores 11

networks 12

resource pools 10

See also deleting resource contention, addressing 43 resource pools

adding 10

removing 10

viewing 55 resource usage 45 resourcepool add command 10, 94 resourcepool commands 94 resourcepool delete command 10, 95 resourcepool list command 10, 55, 95 resources, contention and number of nodes 47

running

HDFS commands 57

Hive scripts 59

MapReduce jobs 58

Pig scripts 58

S

scaling

clusters 42, 78

CPU 43

manual 45

parameters, configuring 48, 78

parameters, resetting 77

RAM 43 scheduling, fixed elastic scaling 50 Serengeti servers

connecting to 80

disconnecting from 82 Serengeti service, connecting 7

Single Sign-On (SSO) 7 starting, clusters 42, 79

stopping, clusters 42, 79

T

topology

adding 33 cluster 31 commands 95 constraints 34

placement policies 32 topology upload command 33, 95 topology list command 32, 33, 95

U

upgrading clusters 80 uploading, topology 33

V

viewing

clusters 77

datastores 54, 81

Hadoop distributions, available 53, 82

networks 54

provisioned clusters 54

resource pools 55 virtual machines, predefined sizes for

Serengeti 66 vSphere Fault Tolerance (FT) 51 vSphere resources

assigning to clusters 20 managing 9 resource pools 10

vSphere High Availability (HA) 37, 51

W

white listed Hadoop attributes 68 worker nodes for Hadoop 17 workloads, balancing 33

X

XML Hadoop configuration, converting to

JSON 70

100 VMware, Inc.

VMware vSphere Big Data Extensions - 2.0 User’s Manual

Specifications and Main Features

Frequently Asked Questions

User Manual

Contents

About This Book

Access the Serengeti CLI By Using the Remote Command-Line Interface Client

Add a Resource Pool with the Serengeti Command-Line Interface

Remove a Resource Pool with the Serengeti Command-Line Interface

Add a Datastore with the Serengeti Command-Line Interface

Remove a Datastore with the Serengeti Command-Line Interface

Add a Network with the Serengeti Command-Line Interface

Reconfigure a Static IP Network with the Serengeti Command-Line Interface

Remove a Network with the Serengeti Command-Line Interface

Creating Hadoop and HBase Clusters 3

About Hadoop and HBase Cluster Deployment Types

Serengeti’s Default Hadoop Cluster Configuration

Create a Default Serengeti Hadoop Cluster with the Serengeti Command-Line Interface

Create a Cluster with a Custom Administrator Password with the Serengeti Command-Line Interface

Create a Cluster with an Available Distribution with the Serengeti Command-Line Interface

Create a Hadoop Cluster with Assigned Resources with the Serengeti Command-Line Interface

Create a Customized Hadoop or HBase Cluster with the Serengeti Command-Line Interface

Create a Hadoop Cluster with Any Number of Master, Worker, and Client Nodes

Create a Data-Compute Separated Cluster with No Node Placement Constraints

Create a Data-Compute Separated Cluster with Placement Policy Constraints

Create a Compute-Only Cluster with the Serengeti Command-Line Interface

Create a Basic Cluster with the Serengeti Command-Line Interface

About Cluster Topology

Topology Rack-Hosts Mapping File

Topology Placement Policy Definition Files

Create a Cluster with Topology Awareness with the Serengeti Command-Line Interface

Create a Data-Compute Separated Cluster with Topology Awareness and Placement Constraints

Serengeti’s Default HBase Cluster Configuration

Create a Default HBase Cluster with the Serengeti Command-Line Interface

Create an HBase Cluster with vSphere HA Protection with the Serengeti Command-Line Interface

Stop and Start a Hadoop or HBase Cluster with the Serengeti Command-Line Interface

Scale CPU and RAM with the Serengeti Command-Line Interface

Reconfigure a Hadoop or HBase Cluster with the Serengeti Command-Line Interface

About Resource Usage and Elastic Scaling

Enable Elastic Scaling for a Hadoop Cluster with the Serengeti Command-Line Interface

Enable Manual Scaling for a Hadoop Cluster with the Serengeti Command-Line Interface

Configure Scaling Parameters with the Serengeti Command-Line Interface

Schedule Fixed Elastic Scaling for a Hadoop Cluster

Delete a Hadoop or HBase Cluster with the Serengeti Command-Line Interface

About vSphere High Availability and vSphere Fault Tolerance

Reconfigure a Node Group with the Serengeti Command-Line Interface

Recover from Disk Failure with the Serengeti Command-Line Interface Client

View Available Hadoop Distributions with the Serengeti Command-Line Interface on page 53

View Provisioned Hadoop and HBase Clusters with the Serengeti Command-Line Interface

View Datastores with the Serengeti Command-Line Interface

View Networks with the Serengeti Command-Line Interface

View Resource Pools with the Serengeti Command-Line Interface

Run HDFS Commands with the Serengeti Command-Line Interface

Run MapReduce Jobs with the Serengeti Command-Line Interface

Run Pig and PigLatin Scripts with the Serengeti Command-Line Interface

Run Hive and Hive Query Language Scripts with the Serengeti Command-Line Interface

Cluster Specification Reference 7

Cluster Specification File Requirements

Cluster Definition Requirements

Annotated Cluster Specification File

Cluster Specification Attribute Definitions

White Listed and Black Listed Hadoop Attributes

Convert Hadoop XML Files to Serengeti JSON Files

Serengeti CLI Command Reference 8

cfg Commands

cfg fs Command

cfg info Command

cfg jt Command

cfg load Command

cfg props get Command

cfg props list Command

cfg props set Command

cluster Commands

cluster config Command

cluster create Command

cluster delete Command

cluster export Command

cluster fix Command

cluster list Command

cluster resetParam Command