If your database exceeds the size for a shared database or you have other requirements you can also use a dedicated database setup.
Regarding the setup you can choose from the same options as in the shared database but your cost of the service depends not on the size of the database but on the machines Show archive.org snapshot the cluster runs on.
In order to run a database cluster at least three nodes are necessary. By default we use two database nodes and a so called quorum/arbitrator node. This gives us the necessary guarantees to avoid split-brain scenarios while keeping the costs for machines low as the quorum node can be of smaller size.
We run a
Galera MariaDB cluster in a synchronous Primary-Primary setup with redundancy and automatic fail-over.
We run PostgreSQL with
etcd in a Primary-Replica setup with redundancy and automatic fail-over.
A dedicated cluster it's completely managed by us, so there are no keys/users configured for ssh access (in contrast to the deploy users on application servers). Let us know if you need special configurations.
Regarding the deployments you can therefore follow the normal configuration.