Installation Prerequisites

Lists the prerequisites for HPE Ezmeral Unified Analytics Software installation on user-provided hosts in connected (internet access) and air-gapped (no internet access) environments.

You can install HPE Ezmeral Unified Analytics Software on user-provided hosts. User-provided hosts are machines that meet the prerequisite criteria listed on this page. A user-provided host is a bare-metal machine or virtual machine (VM) that meets the documented prerequisites. HPE Ezmeral Unified Analytics Software supports bare-metal and VM installations on AWS, GCP, and Azure.

If you need to add additional user-provided hosts to increase the amount of resources for applications and users after you install HPE Ezmeral Unified Analytics Software, you can expand the cluster, as described in Expanding the Cluster.

IMPORTANT
  • The HPE Ezmeral Unified Analytics Software product downloaded includes the software binaries, installation script, Air Gap Utility (for air-gapped environments), and a README.txt file. After purchasing HPE Ezmeral Unified Analytics Software, the downloads are made available to you through the Access your products button in the HPE Subscription Electronic Receipt email that you receive from HPE.
  • When creating a domain name, opt for a subdomain name that is only used for HPE Ezmeral Unified Analytics Software. For example, if your top level domain (TLD) name is company.com, use a subdomain name such as ezua.company.com.

Software Binaries

The README.txt file included with the product provides instructions for downloading and extracting the HPE Ezmeral Unified Analytics Software binaries that are required to install the product, including the Air Gap Utility.

Air Gap Utility (Required only for air-gapped environments)

Use the Air Gap Utility to get the required container images. Create a local repository if you do not already have one. In addition to setting up a local repository, you must also set up a RHEL/Rocky 8-based yum repository. The installer runs yum commands against this repository. See Using the Air Gap Utility for additional information. For operating system support, see the Operating System support matrix.

Host Machines

HPE Ezmeral Unified Analytics Software installation requires two types of host machines with the following minimum storage requirements:
Table 1. Host Machine Requirements
Machine Type VCPU Memory Size (GB) Disk Size (Disk Count) Machine Count
Control plane (HPE Ezmeral Coordinator/Management Cluster) 4 32 500 (1) 2
Workload 32 128 500 (2) 3
Launcher Host1 2 4
1 This is the host that runs the installation script. This host is separate from the hosts that deploy HPE Ezmeral Unified Analytics Software. This host must also have Docker version 20.10 (with a minimum of 20GB storage) and a firewall allowed on port 8080.
IMPORTANT
  • If you plan to use GPU nodes as storage-capable hosts to meet the vCPU sizing requirements, contact HPE Support.
  • Host machines must have a sudo password.
Mount Points
The supported mount points and their minimum sizes depend on the type of host. The host file system must have at least the root mount point: /

The total minimum required size is the combination of all the mount point sizes listed in the table for a given type of host. If you choose not to configure a listed mount point, that mount point's required size must be added to the root (/) mount point.

For example, if you choose not to configure /opt as a separate mount point on the Controller host, you must add the 100GB listed for /opt to the 50GB listed for the root mount point (/). That is, if /opt is not a separate mount point, the Controller host requires 150GB for the root (/) mount point.

Table 2. Mount Point Requirements for Kubernetes Hosts
Mount Point Minimum Size (GB) Purpose
/ 70 Root file system where the Unified Analytics components are stored.
/var, or /var/lib, or /var/lib/containerd , or /var/lib/docker 150 Stores container metadata information. /var/lib/containerd is used for hosts running the Hewlett Packard Enterprise distribution of Kubernetes. /var/lib/docker is used for the other hosts in the deployment.
/opt 50 Stores all Unified Analytics software. /opt/ezkube (on Kubernetes hosts hosts only), /opt/bluedata, and /opt/hpe are used to install Unified Analytics.
VCPU Sizing Guidelines
  • The combined VCPUs of worker hosts should be no less than 96 for deploying HPE Ezmeral Unified Analytics Software services and apps; otherwise, some services cannot start due to lack of resources. If the machine configuration has 16 VCPU and 64 GB of memory, HPE recommends using a minimum of six (6) machines.
  • When installing HPE Ezmeral Unified Analytics Software, the VCPU option (on the Installation Details screen) should be equal to or less than the total VCPUs of the combined capability of the worker hosts; otherwise, the installation will fail due to lack of resources.
DNS Configuration
The DNS configuration requires that:
  • All hosts have A records added to DNS.
  • The name resolution works forward and backward.
  • The FQDN is a maximum of 63 characters.
  • All hosts must be part of the DNS domain and be able to resolve the FQDNs.

Primary Storage

Starting in HPE Ezmeral Unified Analytics Software version 1.5, Unified Analytics no longer provides an internal data fabric for storage on worker nodes. Instead, Unified Analytics supports connections to a primary storage platform that you provide and configure. Primary storage is an external storage platform that Unified Analytics uses to store and process the data generated by the applications and frameworks that run in your Unified Analytics cluster. Unified Analytics connects to the external storage platform through a CSI driver. During Unified Analytics installation, you provide the information that the CSI driver needs to make the connection to the external storage platform. Currently, Unified Analytics supports HPE Ezmeral Data Fabric as primary storage.

Before you install Unified Analytics, the HPE Ezmeral Data Fabric administrator must prepare the HPE Ezmeral Data Fabric cluster. For instructions, see Preparing HPE Ezmeral Data Fabric to be Primary Storage for HPE Ezmeral Unified Analytics Software.

Object Storage

Starting in HPE Ezmeral Unified Analytics Software version 1.5, Unified Analytics no longer provides internal object storage. Instead, Unified Analytics supports connections to S3-compatible object storage that you provide and configure during installation. Currently, Unified Analytics supports the following S3-compatible object stores:
  • HPE Ezmeral Data Fabric Object Store
  • HPE GreenLake for File Storage
  • AWS S3
  • MinIO
Unified Analytics must have permission to create buckets, read from the bucks, and write to the buckets. When configuring object storage during installation, note the following points:
  • For secure object stores (https), specify the FQDN in the object store endpoint TLS certificate. For example, if the object store is located at 192.168.0.10 and uses a TLS certificate with the subject name myobjecstore.example.com, you would specify the endpoint as:
    https://myobjecstore.example.com:9000
    Specifying https://192.168.0.10:9000 would cause a connection failure.
  • You must provide the CA certificate for the object store. Unified Analytics validates the TLS certificate presented by the object store endpoint. If the object storage is within the same primary storage platform, such as HPE Ezmeral Data Fabric and HPE Ezmeral Data Fabric Object Store, use the same CA file that was used for the primary storage.
  • The access key and secret key must correspond to an IAM account on the object store that has permission to create buckets, read from buckets, and write to buckets on the object store. For example, the following IAM policy is acceptable:
    {
  "Version": "2012-10-17",
  "Statement": [
    {
      "Sid": "AllAccess",
      "Action": "s3:*",
      "Effect": "Allow",
      "Resource": [
         "arn:aws:s3:::*",
      ],
      "Principal": "*"
    }
  ]
}

Operating System

HPE Ezmeral Unified Analytics Software supports RHEL 8.8. All machines serving as hosts must run the same OS. Both GPU and non-GPU hosts are supported with RHEL 8.8.

Hewlett Packard Enterprise strongly recommends using only dedicated hosts with clean OS installations on them. Installing HPE Ezmeral Unified Analytics Software on hosts with other running applications can cause unpredictable behavior. To ensure your OS has the latest packages, Hewlett Packard Enterprise recommends performing a yum update before installation.

Use the standard OS kernel; modifications can cause HPE Ezmeral Unified Analytics Software to function unpredictably.

To minimize the need for troubleshooting, Hewlett Packard Enterprise recommends newer kernel versions.

HPE Ezmeral Unified Analytics Software does not support upgrades between major OS versions. For example, if you are migrating from OS version 7.x to 8.x, you must perform a new installation (not an upgrade), and then install HPE Ezmeral Unified Analytics Software.
RHEL 8.8 Requirements
HPE Ezmeral Unified Analytics Software has the following RHEL requirements:
  • RHEL systems must have active, valid subscriptions in order to access the RHEL RPM repositories.
  • Firewall is supported only in iptables mode for RHEL 8.8.
TIP
  • The GPU operator does not support Ubuntu or Rocky.
  • Rocky works in non-GPU environments.
  • For information related to operating systems and operating-system version support, contact HPE Support.

Network

HPE Ezmeral Unified Analytics Software has the following network requirements:
Table 3. Network Requirements
VM DNS Network IP Address
Single vNIC, static IP DNS server to resolve the FQDN Single network connecting all machines. Static: 2 controller machine + 3 worker machines

Port Access

See Ports Used by HPE Ezmeral Unified Analytics Software.

GPU (Optional)

HPE Ezmeral Unified Analytics Software supports GPU on user-provided hosts. For GPU configuration information, see GPU Support.