VR replication with a thin provisioned server

[dbutch1976]

Hello,

I have configured replication for a 1.5 TB file server which is thinly provision and is only using 70GB presently.  It is taking an extremely long time.  As a point of comparison I have configured replication for a 700GB thick provisioned lazy zeroed Exchange server and it completed successfully in about 6 hours.

My question is this:

Does replication actually replicate 1.5 TB of data even when only 70GB is being used?  I was under the impression that using thin provision would greatly reduce replication times of these VMs and with this in mind we have aggressively converted most of our VMs to thin provisioned, however I am seeing no improvement in replication times.

Thanks,

Duncan.

[mvalkanov · Accepted solution]

Hi,

vSphere Replication reads the whole file, computes checksum per each block and exchanges only different blocks.

The all-zeroes blocks are not replicated.

However, with current releases, even the unallocated blocks are read and checksum is calculated for them. This is a known issue and is on the roadmap for a future release.

Regards,

Martin

[MattG]

How does vSphere Replication handle initial replication of thick provisioned servers?   If I have a 1TB thick provisioned VM that is only using 100GB,  will it replicate the 900GB of unused blocks?   Also,  is the latest version of vR more efficient with calculating the empty blocks?

Thanks,

-MattG