What's New in Release 8.2

Following are some new features and capabilities that distinguish release 8.2.0 from the previous release (8.1.0):

Migration of Existing S3 Buckets to Inline Erasure Coding (IEC)

Data Fabric 8.2.0 introduces the ability to convert existing scheduled Erasure Coding (EC) S3 buckets to Inline Erasure Coding (IEC) in place, eliminating the need to delete, recreate, or re-ingest data into new buckets.

Once migrated, incoming object writes are erasure-coded and protected inline during ingestion rather than relying on delayed scheduled tiering and compaction jobs. Administrators can perform and coordinate the migration using the mossdb manage migrateToInstantEC command along with standard volume offload verification workflows.

For complete migration prerequisites, shared-volume constraints, and step-by-step instructions, see Migrating S3 buckets to Inline Erasure Coding (IEC) buckets.

Air-Gapped Repository for Data Fabric Installation

The Data Fabric air‑gapped repository is a package repository that is hosted on a network with no internet access (an air‑gapped environment), so that you can install or update Data Fabric software without connecting to the internet.

You can install Data Fabric by creating an air-gapped repository on a local mirror that hosts the Data Fabric packages offline. You can use the air-gapped repository to download Data Fabric packages for Data Fabric installation, without having to connect to the internet.

You would need to connect to the internet only to update the packages hosted on air-gapped repository with the latest Data Fabric patches and release versions, whenever they are made available by HPE.

See Installing with the Air-Gapped Data Fabric Repository for details.

Object Copier Hub Enhancements

Server-Side ACE and PBS Access Checks for NFSv4

Data Fabric 8.2.0 introduces support for server-side permission enforcement over NFSv4 for resources governed by MapR Access Control Entries (ACEs / ACLs) or Policy-Based Security (PBS).

By enabling the new support_server_aces configuration setting in NFSv4, local metadata cache (mdcache) access checks are bypassed in favor of FSAL server-side test access. This ensures that all permission validations for files and directories configured with MapR ACEs or PBS policies are executed directly on the MapR File System (MFS) server. See Configuring the NFSv4 Server for details.

Spark MCP Server Multi-catalog Support

Data Fabric 8.2 introduces multi-catalog support in the Spark Model Context Protocol (MCP) server. Administrators can now configure and query multiple Iceberg catalogs simultaneously—including both Polaris and Hive metastores—within a single Spark MCP server instance. See Data Fabric MCP Server Settings for details.

Agentic AI and MCP Integration for Data Catalog and Airflow

Data Fabric 8.2.0 extends the Model Context Protocol (MCP) server to integrate with Apache Polaris data catalogs and Apache Airflow orchestration. AI agents and MCP-compliant clients (such as Claude Code) can discover Iceberg tables, execute Spark SQL queries against MapR-FS data, and inspect, trigger, or safely deploy Airflow DAGs using tokenized review workflows. All interactions leverage Keycloak SSO and OAuth 2.0 PKCE to enforce end-to-end user identity propagation and cluster security policies.

See Airflow MCP Integration for Agentic AI for configuration and usage details.

Data Fabric UI Enhancements

Important features introduced in the earlier releases

The major features introduced in the earlier Data Fabric releases are listed below for your reference.

The object lifecycle can be used to manage the expiration and auto-deletion of expired versioned and non-versioned objects, that reside on the object store buckets. See Object Lifecycle for details about the object lifecycle.

See mc ilm to manage object lifecycle management rules by using the CLI.

See mc admin ilm-scanner to start, stop, monitor, and control scan throttling.

You can copy objects across fabrics or from AWS S3 buckets to Data Fabric buckets using the Object Copier Hub.

The Object Copier Hub is a service that must be installed on the node from which you wish to initiate the copying of objects. You can initiate the copy operation from a node that is not a part of a fabric created with Data Fabric.

The Object Copier Hub facilitates the addition of the source bucket - destination bucket pairs so that objects can be copied from the source bucket onto the destination bucket.

When the copy operation is be carried out across fabrics, the source bucket would belong to the S3 object store on one fabric and the destination bucket would belong to the S3 object store on another fabric.

When the copy operation is be carried out from an AWS S3 bucket to a bucket on a fabric created with Data Fabric, the source bucket would belong to AWS S3 and the destination bucket would belong to the S3 object store on the fabric.

Inline erasure coding or instant tiering is a new feature that facilitates the backing up of data from front-end volumes is to backend volumes without the need for a schedule. The instant tiering process is a continuous one and helps with storage optimization as against the traditional scheduled tiering that has been available in HPE Data Fabric from the 6.2 release.

For information on inline erasure coding or instant tiering, see Inline Erasure Coding.

To understand the differences between non-line erasure coding (scheduled tiering) and inline erasure coding (instant tiering), see Non-inline Erasure Coding Vs Inline Erasure Coding.

The Flink SQL client can be used for sourcing and sinking data between Data Fabric Streams and Iceberg tables managed by DEP Polaris. See Flink SQL Client and Iceberg Integration Demofor an integration demo.

Data Fabric 8.1.0 adds Polaris integration with Global Namespace (GNS), enabling you to run Spark analytics across multiple clusters, accessing Polaris catalogs and data on remote clusters through external storage and Polaris Federation, without moving data. See Polaris and Global Namespace Integration Support for more information.

Data Fabric 8.1.0 introduces Polaris MCP integration, enabling you to configure Polaris as a metastore for Apache Spark and expose MCP endpoints for streamlined data catalog access across your Data Fabric environment. See Configuring Polaris as a metastore for Spark for more information.

Beginning with release 7.10.0 and EEP 9.4.0, librdkafka is now updated to version 2.6.0.

The Data Fabric 7.10 release supports C API Librdkafka 2.6.0 for HPE Data Fabric Streams. See the following topics for details:

• HPE Data Fabric Streams C Applications
• HPE Data Fabric Streams C Client 2.6.0 Release Notes

https://package.ezmeral.hpe.com/releases/v7.8.0/redhat/

https://package.ezmeral.hpe.com/releases/MEP/MEP-9.3.0/redhat/

New Repository for NFS v4 on SLES 15

In release 7.8.0, the /releases directory of package.ezmeral.hpe.com includes two suse repositories:

The suse15.5 repository is new and enables support for NFS or NFS v4 for installations on SLES 15 SP3 and SP5. To use the old and new suse repositories, see this table:

Repository	Use this respository to install	NFS Support
suse	SLES 15 SP2	NFS
suse15.5	SLES 15 SP3 or SP4 or SP5	NFS or NFS v4

New Key for Signature Verification for Data Fabric Files

Release 7.6.1 implemented a new key for .rpm, .tar.gz, .zip, and .tgz files for the following Data Fabric products:

• core 7.6.1 and later
• Data Fabric clients
• Ecosystem Pack (EEP) 9.2.1 and later
• Installer 1.18.0.5 and later

Before you install these products, you must import the HPE GPG public keys. This is a one-time operation on each node where packages are installed. Importing the keys allows you to, optionally, verify the GPG and RPM signatures for the products. A verified GPG or RPM signature attests that the product you received has been signed with digital private keys held only by HPE. Successful signature verification also ensures that the file has not been altered after it was signed and released by HPE.

For more information, see HPE GPG Public Keys for GPG or RPM Signature Verification.