본문 바로가기

Cloud & New Tech

[솔루션] Commvalut Disaster Recovery (CDP)

Reference : https://documentation.commvault.com/disaster_recovery/106449_vmware_continuous_replication_using_vaio.html

 

 

 

Overview

 

The continuous replication feature uses block-level replication to synchronize block devices or virtual machine disks, continuously streaming updates from source to destination VMs.
Each source VM and the resulting destination VM is a replication pair.
vSphere APIs for I/O Filtering (VAIO) is a VMware framework that enables block-level replication of virtual machine data.
You can configure continuous replication only for on-premises VMware instances.
You cannot configure continuous replication for cloud instances, such as VMC, Dell-EMC, Oracle Cloud, and others.

 

** Note: From 11.20, the Command Center uses the name "continuous replication" instead of "Live Sync I/O".

 

Block-level replication provides the following benefits:

  • Eliminates the need to take and clean up snapshots frequently on the source VM, reducing the I/O impact on the production system.
  • Enables more frequent and granular updates and synchronization between source and destination VMs,
    making it possible to meet reduced recovery time objectives (RTOs) and recovery point objectives (RPOs) for virtual machines.
  • Provides a dynamically pruned history of recovery points
    without running backups or relying on synthetic full backups to consolidate incremental updates for different points in time.

You can configure and run continuous replication from Command Center or the CommCell Console.

 

** Note: Continuous replication is not supported in a CommCell environment configured with a Linux CommServe server.

 

Recovery Points

Recovery points are crash consistent copies of volumes on the destination VM that match the data that was written on the source VM volumes at a specific point in time.

Application-consistent backups that are performed with the assistance of application agents or by performing VSA backups create application-consistent recovery points (ACRPs).

ACRPs are based on application-consistent VM snapshots.
If a replication pair is configured for ACRPs, the replication service periodically creates temporary application-consistent snapshots on the source VM.
The Recovery Point Store (RP Store) on the destination site records ACRPs as point-in-time recovery points.

 

 

Architecture for Continuous Replication

Continuous replication attaches Commvault plug-ins to each source VM, leveraging vSphere APIs for I/O Filtering (VAIO filters) to capture block-level I/O for each VM disk.

The Commvault plug-ins for each VM pass transaction records through daemons at the source ESX cluster (CVBLRH) to the Block-Level Replication (BLR) service.
The BLR service at the source sends the collected data from the source ESX cluster to the BLR service at the destination cluster, updating the Recovery Point Store (RP Store) and passing updates through CVBLRT daemons to each destination VM.

The RP Store at the destination maintains recovery point data and metadata for each VM.

Commvault Filter Components

 

 

System Requirements for Continuous Replication

This topic summarizes requirements for continuous replication.

To avoid delays in replication, implement the following best practices:

  • Use a low latency network.
  • Provide enough cache space and a fast storage connection to the cache on the access node (VSA proxy) at the source site.
  • Provide a fast storage connection to the RP Store.
  • Provide enough storage for the RP Store.
  • Timestamps must be synchronized between access nodes and ESXi hosts.

Hypervisor Support

vCenter and ESXi Requirements

  • vCenter Server version: 6.5, 6.7, 7.0
  • ESXi version: 6.0, 6.5, 6.7, 7.0
  • ESXi memory requirement for each VM: 100MB
  • On the ESX cluster, vSphere High Availability (vSphere HA) must be configured and vSphere Distributed Resource Scheduler (DRS) must be enabled.

Hardware

  • Two access nodes (VSA proxies): One at the source site and one at the destination site.
  • MediaAgent should be present on the destination access node.
  • Low latency network between the access nodes, preferably 10 GB ethernet, depending onI/O loads.
  • The destination requires a fast storage connection to the Recovery Point Store (RP Store).

Virtual Machine Disks

Before configuring a source VM for continuous replication, ensure that the VM storage policy in vSphere is set to Datastore Default for all disks.

Access Node Operating Systems

You can deploy access nodes (VSA proxies) on physical or virtual machines running the following operating systems:

  • Microsoft Windows Server 2019
  • Microsoft Windows Server 2016
  • Microsoft Windows Server 2012 R2 x64 Editions
  • Microsoft Windows Server 2012 x64 Editions
  • Microsoft Windows Server 2008 R2 x64 Editions

Note: Linux access nodes are not supported.

Plug-In Installation or Upgrade Requirements

VMware vCenter Server Privileges

  • Ensure that vCenter permissions are configured correctly as described in Permissions for Custom User Accounts.
  • The VMware vCenter Server user account used to install plug-ins must have administrator privileges for installation or upgrade.

Acceptance Level

  • For each ESX host, the acceptance level can be one of the following:
    • VMwareAccepted
    • PartnerSupported
    • CommunitySupported

Licenses and Features Enabled

  • vSphere Enterprise Plus license is required to support continuous replication.
  • During installation or upgrade of the Commvault plug-in, ESX servers go into maintenance mode, which requires that vSphere High Availability (vSphere HA) is configured and vSphere Distributed Resource Scheduler (DRS) is enabled.
  • Plug-ins must be part of protected and replicated data centers (ESX clusters).

Source Access Node Cache Requirement

  • Cache location: Job Results directory

Network Bandwidth

Provide network bandwidth based on the estimated I/O speed. For example:

  • Protected VMs: 10
  • Rate at which each VM performs I/O: 100 MB/s
  • Preferred Network Speed: 1000 MB/s

Preferred Recovery Point Store (RP Store) Requirement

Size the RP Store based on I/O throughput and RP Retention criteria. For example:

  • RP Store I/O throughput: 100 MB/s
  • Total Retention Time (hours): 5 Hrs
  • RP Store size: 1800000 MB (1.72 TB)

Firewall Port Requirements

For environments that include a firewall, open the following ports:

  • For ESXi servers on the source site, open port 33040.
  • For ESXi servers on the destination site, open port 33041.

These ports are used by the BLR service to communicate with VAIO daemons.