Hi @dario72  doesn't make sense, rather if it is a mission critical server, you can enable the Vmware “Fault tolerance” feature that is also supported for Vsan.
 

Best Practices for Fault Tolerance (vmware.com)

regards

It does not make sense but can be done as you need to pin the CDP proxies to specific hosts to have it work. I have done this in homelab without issue but considering everything is same cluster it does not make sense as mentioned.  If it was DR site then sure.

As @Chris.Childerhose and @Link State said, it doesn’t make sense from a site redundancy perspective.

I think it might well make sense from a workload perspective 

CDP provides the lowest possible RPO, even down to a few seconds. 

Consider a file server that is encrypted by ransomware. With a CDP policy replicating it to another VM, even in the same cluster, you would be able to start the VM in the state it was in just seconds before disaster.

Interesting to know the use case… What is the justification to need this?  Why are other features (FT in VMware) not suitable? 

the use case is: the customer has a Microsoft SQL Server and does not want to have a failover cluster, so he wants to have CDP so that he can switch over quickly if a problem occurs, and he wants to go back over the short-term retention, 4 hours can to go back

the use case is: the customer has a Microsoft SQL Server and does not want to have a failover cluster, so he wants to have CDP so that he can switch over quickly if a problem occurs, and he wants to go back over the short-term retention, 4 hours can to go back

but Veeam say: 

• CDP is not supported for VMs to which VM storage policies with multiple datastore specific rule sets apply. If you need to define datastores to be used for placement of the VMs, you can add a tag rule for vSAN instead of a tag based placement datastore specific rule. For more information on how to create tag rules for vSAN, see VMware Docs.

how did you configure the vsan policy without error tolerance?

I have not used vSAN with CDP so cannot say. I have used it for other testing and pinning the proxies to hosts. 

If the OS messes up, or the DB, this is not as robust as a failover cluster.  Any bad data or errors will be replicated or mirrored. 

Unless it is a license/cost thing, SQL Failover cluster is 100% the right choice.  This may buy some time during a host failure, but even still, with VMware HA/DRS the SQL server will start on the other host. 

You said Cluster so they obviously have vCenter. Do they just have vSphere standard?

With standard (no HA/DRS), and no SQL licenses to make the cluster, then this might make sense, or using ESXi Free.