Implementation of vSAN Cluster and DR Replication using inbuilt per VM replication feature to a offsite location

[waza74]

Hi All,

I'm looking at implementing a 3 node vSan cluster in a production site, and a 2 node vSan cluster in an external datacentre with a latency of <50ms.

For DR I would like to use the inbuilt feature of vSan to replicate 3 or 4 VM's (file/print type servers) to an external data centre, with a RPO of about 8 hours. I dont think I need SRM, just piece of mind knowing I have a very recent replica of these VM's.

I was thinking of installing a 2 node vSan cluster in the data centre, could this be a streched cluster?

I am not very familiar with the vSAN product, but have read some guides available on VMWare website.

Thx

[TheBobkin]

Hello waza74,

First off, welcome to Communities!

Just to clarify a few things

- vSAN doesn't use snapshot or point-in-time replication for data protection - In a stretched cluster with FTT=1 (Failures To Tolerate) using RAID1 as the FTM, VMs data would be accessible at both sites as the data is replicated to both fault domains.

- Stretched-clusters require a RTT latency of <5ms between data sites so maybe this won't be an option here.

- Stretched-clusters require Enterprise version of vSAN licensing (which is required in addition to ESXi licenses).

- It is possible in vSAN 6.6 to have asymmetric clusters (e.g. 3-4 nodes at one site and 2 at the other site) with non-critical VMs being FTT=0 and with data located just at one site and FTT=1 VMs with data replicated to both sites (achieved by defning PFTT and SFTT in Storage Polices).

If site-to-site latency is too high then an alternative would be to set up a second cluster (vSAN or otherwise) and use SRM or a DR solution to store secondary/back-up data.

blogs.vmware.com/virtualblocks/2017/02/09/vsan-srm-rpo-rto/

Bob

[GreatWhiteTec]

waza74,

If all you want is to replicate a few VMs between 2 vSAN clusters without implementing vSAN stretched-cluster, you have the option to use vSphere Replication (VR) for those VMs. Additionally, you can use SRM on top of VR to coordinate the failover between sites.

~Dave