Instant VM Restore: Minimize Downtime with HPE Zerto Software

HPE Zerto Software offers an Instant VM Restore feature, enabling you to recover applications in seconds with minimal performance impact, ensuring your business remains resilient and continuously operational.

Built on Continuous Data Protection (CDP), this enables near-zero data loss and downtime, measured in minutes, allowing for the rapid failover and recovery of entire virtual machines (VMs) or multi-VM applications.

In the recovery tests presented below, it was possible to restore a VM almost immediately, with Recovery Point Objectives (RPO) as low as 5 seconds and Recovery Time Objectives (RTO) as low as 16 seconds, all with minimal impact on performance during the recovery process.

Local Continuous Backup

HPE Zerto Software Local Continuous Backup is a data protection solution that provides continuous, journal-based backup of virtual machines and applications, creating recovery points every few seconds in the same environment.

Unlike traditional backup methods that rely on periodic snapshots taken every 4, 8, or 24 hours, HPE Zerto continuously captures every write change made by the application through the Zerto Virtual Replication Appliance (VRA) and immediately copies it to the local journal at the production site’s datastore.

At regular intervals of a few seconds, the solution creates a checkpoint timestamp across all journals to ensure write-order fidelity and crash consistency. This process preserves the exact sequence of write operations as they occurred, enabling recovery to precise and consistent points in time. Each checkpoint is assigned a unified timestamp, which is managed and synchronized by the Zerto Virtual Manager to maintain consistency across the applications.

The local replica disk serves as the baseline recovery target for the protected VM and is also located in the datastore of the protection site. As the journal accumulates data over a pre-defined period, older writes and checkpoints are periodically flushed from the journal to the replica disk.

This process permanently commits those changes, ensuring that the replica disk consistently reflects a stable state of the protected VM disk.

In the figure above, we have some parameters.

A virtual SQL server is replicating locally to the journal, with an RPO of 05 seconds. In addition to the journal, a replica disk is also being stored locally and associated with this VM.

• SQL server VMDK disk = approximately 8.2 GB
• Replica disk = approximately 7.4 GB
• Journal = approximately 1.3 GB

To simulate a failure on the virtual SQL database server, a “disk detach” action was performed in the VMware vCenter console, followed by a power-off.

In this way, after this action, we rely solely on the local journal and the replica disk to recover the operation of this VM.

The recovery operation is effortless and fast.

In the Zerto Virtual Manager (ZVM) console, we select the SQL VM we want to restore, choose the latest recovery point (checkpoint) before the “disk detach action”, and select the “Instant Restore” operation.

A new VM SQL (1) is created immediately. The restored VM is a replica of the original VM’s state at the selected checkpoint within the local journal. This process does not overwrite or affect the original SQL VM or its protection; instead, it generates a cloned VM that operates independently from the original.

The cloning process enables you to quickly resume operations without waiting for a complete data copy from backup storage to production storage.

After that, HPE Zerto presents the replica disks and journal disks together as virtual disks attached to the cloned VM SQL (1).

Because the new VM runs directly from a combination of replica disks (containing the baseline data) and journal disks (storing recent writes) that are immediately ready to be attached, it enables near-instant VM availability.

If the “Auto-commit” option with a duration of “0” seconds is enabled during the operation in the Zerto Virtual Manager (ZVM) console, the promotion process begins automatically.

The cloned VM SQL (1) is then promoted to become the new production VM. It is registered in the production inventory and assumes either the original VM’s identity or a new identity as configured.

Once committed, the VM SQL (1) is fully operational and ready for use.

The measured recovery performance from ZVM and vCenter was:

• RTO = 16 seconds
• RPO = 05 seconds

While the restored VM is running, all new writes it generates are initially written to the journal, preserving write-order fidelity and ensuring crash consistency. The journal functions as a write cache overlay, temporarily storing recent writes to disk.

Gradually, the data is flushed from the journal to the replica disk, updating the baseline VM data. This process is known as promotion.

During this process, while the VM operates using both the journal and replica disks, the associated Virtual Protection Group (VPG) remains in the “Instant VM Restore” state in the Zerto Virtual Manager console.

After the promotion process is complete, the associated VPG status returns to the “normal state” in the ZVM console.

The restored VM runs from the replica disk, which becomes its active storage. The original replica and journal data remain available in the background to support additional recoveries or rollbacks if necessary.

In the figure above, we can verify that the original journal was preserved: “180_DS-VM_Left_vm-8051”.

Preserving journal checkpoints enables reverting to an alternative point in time for the same VM if the initially selected restore point does not meet the requirements.

After that, we can edit the VPG and change the protected VM from VM SQL to VM SQL (1). A new journal and replica disk are created for this VM SQL (1).

It´s also possible to keep the previous journal and replica disks in the ZVM console, if necessary.

This is a simple yet powerful process that minimizes RTO, RPO, and performance impact because both the journal and replica are stored locally, typically on production storage, and ready to be attached to the recovered VM.

As a result, this approach from HPE Zerto Software is ideal for fast local recovery with minimal data loss, especially for critical applications.

References:

Instant VM Restore from the Journal

Hewlett Packard Enterprise (HPE) – Auth – Sign In

Demo- HPE Zerto Software in 2 minutes