Do we have an option to use disks for persistent hbase data instead of using ADLS or azure storage for HDINSIGHT hbase cluster?

Question

Do we have an option to use disks for persistent hbase data instead of using ADLS or azure storage for HDINSIGHT hbase cluster?

Karthik Gummadi 0

Pilladi Padma Sai Manisha 4,990 Reputation points Microsoft External Staff Moderator

2026-03-02T21:29:32.89+00:00

Hi Karthik Gummadi,
Thankyou for reaching microsoft Q&A!
In Azure HDInsight HBase, there is no supported option to use VM-attached disks as persistent storage for HBase data instead of Azure Storage (ADLS Gen2 or Blob Storage).

HDInsight is architected with a separation of compute and storage. All HBase data (including HFiles and WAL) is stored in Azure Storage, which acts as the persistent layer. This design ensures durability, scalability, and allows the cluster to be safely deleted and recreated without data loss.

The local disks attached to the cluster VMs are temporary and tied to the lifecycle of the nodes. Any data stored on those disks would be lost if the cluster is deleted, resized, or the nodes are reimaged. Because of this, HDInsight does not support using local or managed disks as the primary persistent storage for HBase.

If you require a disk-based persistent storage model similar to traditional on-premises HDFS deployments, you may need to consider alternative architectures outside of HDInsight, since the service is designed to rely on Azure Storage for durability.
Manoj Kumar Boyini 8,950 Reputation points Microsoft External Staff Moderator

2026-03-03T19:05:16.2033333+00:00

Hi Karthik Gummadi,

I hope you had a chance to review the information shared earlier, and I hope this information has been helpful! If you still have questions, please let us know what is needed in the comments so the question can be answered.

1 answer

Your answer

Pilladi Padma Sai Manisha 4,990 Reputation points Microsoft External Staff Moderator

2026-03-02T21:29:32.89+00:00

Hi Karthik Gummadi,
Thankyou for reaching microsoft Q&A!
In Azure HDInsight HBase, there is no supported option to use VM-attached disks as persistent storage for HBase data instead of Azure Storage (ADLS Gen2 or Blob Storage).

HDInsight is architected with a separation of compute and storage. All HBase data (including HFiles and WAL) is stored in Azure Storage, which acts as the persistent layer. This design ensures durability, scalability, and allows the cluster to be safely deleted and recreated without data loss.

The local disks attached to the cluster VMs are temporary and tied to the lifecycle of the nodes. Any data stored on those disks would be lost if the cluster is deleted, resized, or the nodes are reimaged. Because of this, HDInsight does not support using local or managed disks as the primary persistent storage for HBase.

If you require a disk-based persistent storage model similar to traditional on-premises HDFS deployments, you may need to consider alternative architectures outside of HDInsight, since the service is designed to rely on Azure Storage for durability.
Manoj Kumar Boyini 8,950 Reputation points Microsoft External Staff Moderator

2026-03-03T19:05:16.2033333+00:00

Hi Karthik Gummadi,

I hope you had a chance to review the information shared earlier, and I hope this information has been helpful! If you still have questions, please let us know what is needed in the comments so the question can be answered.

Answer 1

Yes, you can use disks for persistent HBase data in an Azure HDInsight cluster. The Accelerated Writes feature allows you to attach premium SSD-managed disks to every RegionServer (worker node). This means that Write Ahead Logs are written to the Hadoop File System (HDFS) mounted on these premium managed disks instead of using Azure Data Lake Storage (ADLS) or Azure Storage. This setup provides low write-latency and better resiliency for your applications.

References:

Share via

Do we have an option to use disks for persistent hbase data instead of using ADLS or azure storage for HDINSIGHT hbase cluster?

1 answer

Your answer