Total Pageviews

My YouTube Channel

Tuesday, 8 January 2013

vSphere Replication FAQ


This article provides information about vSphere Replication, a feature in VMware vCenter Site Recovery Manager (SRM) 5.0.


What is vSphere Replication?

vSphere Replication is a feature of Site Recovery Manager (SRM) that allows SRM to Disaster Recovery (DR) protect individual virtual machines without requiring array-based replication. This feature is licensed as part of the SRM product.

What are the new use cases enabled for SRM by vSphere Replication?

vSphere Replication:
  • Allows per virtual machine replication for SRM
  • Can use SRM without requiring array-based replication
  • Enables virtual machine replication between heterogeneous arrays
In addition, SRM can be deployed to DR protect virtual machines regardless of which Hardware Compatibility List (HCL) supported storage is used at primary or recovery sites.

Where is the Getting Started help?

The Getting Started help screen, which helps you configure vSphere Replication and SRM, is hidden from most users. To see this screen, click Edit > Client Settings in the vSphere Client and select Show Getting Started Tabs.

How do I enable vSphere Replication?

vSphere Replication can be enabled when SRM installation is complete. You must install and configure vSphere Replication from within the SRM plug-in before actually replicating a virtual machine. Replication is actually enabled in the Host/Clusters view within the vSphere Client.

What are the minimum requirements for enabling vSphere Replication?

To enable vSphere Replication, you must use:
  • SRM 5.0
  • ESXi 5.0
  • vCenter Server 5.0
  • Virtual machine hardware version 7

What are the roles of the various vSphere Replication system components?

  • vSphere Replication Agent and vSCSI Filter Drive – These are ESXi components (which ship with ESXi) and they manage the replication process. They provide these functions:
    • Scheduling replications
    • Transferring data to remote vSphere Replication servers
    • Coordinating replication of virtual machine configuration, and group consistency for virtual machine disks
  • vSphere Replication Management Server (VRMS) – This is the generic management framework for vSphere Replication. It is a Linux-based virtual appliance, with one VRMS server per vCenter Server, and it is deployed, configured, and managed by SRM. It provides these functions:
    • Authentication / authorization for vSphere Replication
    • Mapping between protected and replica disks
    • Allocating storage resources at the recovery site
    • Monitoring vCenter for changes that may impact replication
    • Orchestrating creation of test and failover images
    • Providing vSphere Replication support to SRM
  • vSphere Replication Server (VRS) – This is the component that receives replication traffic from the protection site. It is a Linux-based virtual appliance, with one or more deployed at the recovery site. It is deployed, configured, and managed by SRM.

Do the source and destination datastores need to be of the same type for using vSphere Replication?

No. In fact, there might not even be a VMFS backing the virtual disks. Virtual disks can be stored on NFS shares or even raw LUNs.

If using VMFS, do the source and destination volumes for vSphere Replication require the same block size?

No. vSphere Replication deals with disks at a much higher level, and is unaware of the block sizes that the underlying VMFS is using to store VMDKs.

Does the vSphere Replication feature incorporate WAN bandwidth management?

No. vSphere Replication does not incorporate WAN bandwidth management. Use WAN optimizers to throttle and prioritize vSphere Replication traffic. To make it easier, vSphere Replication separates the initial replication from the ongoing replication using different ports so that you can assign different traffic shaping rules to the initial replication and ongoing replication.

How does vSphere Replication manage bandwidth consumption for initial and on-going replication?

vSphere Replication uses different TCP ports for full and on-going replication.
The port numbers are:
  • Port 31031 for initial replication
  • Port 44046 for on-going replication

Can vSphere Replication tasks be scheduled?

No. vSphere Replication is Recovery Point Objective (RPO) based and scheduled per ESXi host. 15 minutes is the most aggressive RPO that can be selected via the UI.

Can powered off virtual machines be replicated using vSphere Replication?

Powered off virtual machines can be configured for vSphere Replication. However, actual replication of data occurs when the virtual machine is powered on.

Does vSphere Replication maintain virtual machine snapshot hierarchy?

No. vSphere Replication does not maintain virtual machine snapshot hierarchy at the secondary site.

What are some other differences between array-based replication and vSphere Replication?

This table outlines some differences between array-based replication and vSphere Replication:

FeatureArray ReplicationvSphere Replication
Replicate virtual machine templates
Replicate Fault Tolerance (FT) virtual machinesYes. You have to enable after failover.No
Replicate linked clone treeYesNo
Maintain snapshot hierarchyYesNo
Replicate powered off or suspended virtual machineYesNo, but can configure for replication
VSS application consistency
Possibly. This is array specific, and would include log management.
Yes, but no log management
vApp consistency groupsYesNo
Application consistency AgentsNormally available, sometimes for an extra cost.No

What engine is vSphere Replication using? Is it Changed Block Tracking (CBT) or snapshots?

The engine is designed by VMware. It uses something similar to CBT. vSphere Replication does not use snapshots to replicate.

If data is similar on each of the storage devices (that is, the original copy has already been made), but now the two sides are out of sync, is a new copy made, or are changes sent?

If no changes to the disk have been made "behind the back" of the vSphere Replication infrastructure, then only deltas (sets of changed blocks) are sent to complete a new copy.
If the disks are changed out of purview of the vSphere Replication infrastructure (or if the virtual machine is new to vSphere Replication), the vSphere Replication system compares the disks (without sending the content, just checksums between sites) to re-establish its state, then sends deltas going forward. This complete disk synchronization step involves reading the entire disk at each site, but matching disk blocks are not re-sent.

Are there any special VMDK disk requirements to support vSphere Replication?

No. You do not need eager zeroed thick disks or any other special requirements. Physical RDMs are not supported.

With only the base disk replicated, how are snapshots handled?

Writes are replicated, so vSphere Replication does not know what snapshots are. As such, snapshots are not replicated. When the virtual machine is accessed at the recovery side, it is like the snaphots have been collapsed.

Is there a minimum VMFS format level required for vSphere Replication?

No. All VMFS formats supported by ESXi 5.0 are supported by vSphere Replication.

What databases are supported for VRS? For VRMS?

Only the VRMS server requires a database. It supports both Oracle and Microsoft SQL databases.

Can vSphere Replication be used outside of SRM?

No. vSphere Replication is a feature of SRM, and it cannot be disconnected from SRM.

Does vSphere Replication support MSCS clusters?

vSphere Replication does not work with virtual disks opened in "multi-writer mode". MSCS cluster virtual machines are configured with virtual disks opened in "multi-writer mode," so vSphere Replication will not work with a MSCS configuration.

Does vSphere Replication support Storage DRS?

Neither array-based replication nor vSphere Replication support using Storage DRS. Storage vMotion of a replicated virtual machine results in a full sync, where both the primary and the recovery side disks are read and hashed, and these hashes are exchanged over the wire which can result in heavy I/O, and this can cause latency on the datastore.