Is it possible to use RAFT as the datastore instead of the default consul for replication solution across multiple data centers across regions in enterprise Vault?
The backend shouldn’t make a difference in the distribution model. I’m sure it isn’t best practice nor recommended for a large enterprise though. It probably assumes that in an enterprise a full consul backend is probably better optimized for reads and writes.
Thanks for the reply. Looking for resources/links that describes replication across regions using alternate (preferably RAFT) datastore instead of consul.
Since I believe it isn’t best-practice, at least not yet, in the eyes of Hashicorp I doubt they would publish a document on how to do it. It would be up to you to build your own solution.
I would suggest looking at Vault Reference Architecture | Vault - HashiCorp Learn. The second section describes how to do a single cluster across regions, which as you can see shows that the consul servers won’t communicate in that setup. All replication is done between the vault nodes. If that is how it’s set up then I would guess that using Raft as the storage wouldn’t make a difference.