Step2：Create KMR Cluster

Enter “Basic Information” page

• Cluster name: when the cluster is created, a default name is generated based on the system timestamp. You can enter a custom name for the KMR cluster. The length is limited to 1-25 characters, and it supports numbers, letters and special characters (_ and-). This name does not have to be unique.
• Data center: select the data center where the KMR cluster is located. (If you need to store the data or bind EIP by using KS3, you should ensure that KMR is in the same area as the KS3 bucket or EIP)
• Availability zone: select the availability zone of KMR cluster.
• Product version: select the product version used to create the KMR cluster. Different version can provide different basic cluster configurations and application components. Two versions are supported Currently: KMR3.0.0 and KMR4.0.0. It is recommended to use KMR4.0.0.
• Mandatory components: service components installed by default, which can be managed in Ambari after the cluster is created.
• Optional components: select to install components you need.

  Node group configuration

  

• User quota: when KMR service is activated, a resource quota is allocated for each account. If the cluster resources used in the account exceed this quota, the cluster cannot be created. If you have special needs, please contact your customer manager.
• Master instance: the master instance is used mainly for the cluster management, and it runs the management processes in the Hadoop cluster, such as namenode, resource manager, jobhistory, etc. In addition, it also stores the management component LOG information. You can select to enable Elastic EIP for the master node to access to external network, and MASTER1-2 uses anti-proximity strategy to implement the cluster HA. Master instances do not support big data hosts.
• Core instance: the core instance is used mainly for deploying ZOOKEEPER, Datanode and NodeManager components to execute the storage + computing node.
• Task instance: the task instance is used mainly for the computing, deploying NodeManager, and it does not store HDFS data. It can expand cluster computing resources horizontally. TASK nodes support rapid capacity scale-up/down.
• Gateway instance: the client node will deploy the client operating environment required by the cluster service automatically, and it can perform the job submission and automatic scheduling. Moreover, the user can bind EIP to the client node as the data transmission entrance of the cluster.

  “Network and Others” page is configured

  

• EIP binding: EIP is public network IP address bound to the cluster master node, and it is used mainly for remote cluster management and job submission with the bandwidth of 1Mbps. If you have other requirements, or you want to bind public network EIP after the cluster is created, you can go to EIP console. For the operation performed on EIP console, KMR will synchronize EIP status to KMR page in a period, and during the period that the status is not synchronized, normal use will not be affected.
• VPC network: You can use any VPC in the same data center to create the cluster. If there is no VPC in the data center, you can create it in VPC console, and you can create an EndPoint subnet under VPC and VPC subnet in the same availability zone as the cluster.
• VPC subnet: VPC subnet is a network element in VPC used to manage the cloud host. You can create the cluster by using any VPC subnet under the same VPC as the cluster.
• EndPoint subnet: EndPoint can be used to create a private connection between your VPC and other Kingsoft Cloud services. You must specify EndPoint to use KMR service. You can create the cluster by using any EndPoint subnet under VPC.
• SSH key (optional): if you need to access to the cluster through SSH, you should click "Bind Key" to bind SSH key to the cluster. Please refer to “SSH Key Management”.
• High availability of metadata (optional): this option allows you to configure a highly available metadata repository for the resident cluster. With RDS instance metadata repository, the metadata reliability and read and write performance can be improved. This can be created by selecting RDS instance in the same computer room, and filling in RDS instance port, RDS user name and RDS password.