How Veritas Resiliency Platform Data Mover handles virtual machine writes
Veritas Resiliency Platform Data Mover uses the vtstap service running on ESXi host, to intercept and process protected virtual machine writes. The vtstap service intercepts the writes to the storage, while reads are directly processed from the virtual machine storage.
The vtstap service records the location of the write I/O, and also sequences the writes. The write is applied to the virtual machine storage and sequenced writes are then asynchronously sent to source Replication Gateway. The writes that accumulate on the source Replication Gateway are periodically sent to the target Replication Gateway. The target Replication Gateway applies the writes to the target data center storage in sequence. This ensures that data is consistent on the source and target data centers. As Resiliency Platform Data Mover employs asynchronous replication, there might be a lag between the data on source and target, but it will always be consistent.
Resiliency Platform Data Mover processes an incoming write by performing the following steps in the order listed:
The operating system in the guest virtual machine issues a write to the virtual machine storage.
The I/O tap module (vtstap) records the location of the I/O.
IO is written to virtual machine storage.
The vtstap module sends the I/O data over the network to the I/O receiver in the source Replication Gateway.
The I/O receiver aggregates the I/Os.
Periodically the transceiver on the source Replication Gateway sends the I/Os across the network to the transceiver on the target Replication Gateway.
The I/Os are then sent to the applier and the applier writes them to the target data center storage.
Resiliency Platform supports protecting the virtual machines with thin disks optimally by provisioning corresponding thin disks on the target data center. Resiliency Platform Data Mover replicates only the used blocks from the disks on the source data center to the disks on the target data center thereby maintaining the thin nature of the disks. In addition to thin disks, Resiliency Platform Data Mover also replicates only the used blocks from a thick (lazy zero provisioned) disks to the disks on the target data center.
When you upgrade from 3.1 or an earlier version to version 3.2 or later and if you have thin disks on the source data center and the resiliency group has been configured for disaster recovery (DR), you need to delete the existing resiliency groups after upgrading. Reconfigure the resiliency groups to utilize the thin provisioning feature. If you are upgrading from 3.1 Update 1, you do not need this step.