New to KubeVault? Please start here.
Storage Backends
A VaultServer keeps the data that Vault encrypts and stores in the storage backend set under spec.backend. Exactly one backend can be configured. The guides in this section deploy a VaultServer with each of the supported backends, in the same way as the Raft and PostgreSQL guides: prepare the storage, create the credentials, deploy the VaultServer, then write a secret to see that it works.
To learn what each field of a backend does, read the storage concepts.
Supported Storage Backends
| Backend | High availability | What you need |
|---|---|---|
| Aerospike | No | An Aerospike server with an existing namespace |
| Alibaba Cloud OSS | No | An OSS bucket and an access key pair |
| Amazon DynamoDB | Yes (haEnabled) | AWS credentials (the table is created by Vault) |
| Amazon S3 | No | An S3 bucket and an access key pair |
| Azure Blob Storage | No | An Azure storage account and its access key |
| Cassandra | No | Cassandra with a keyspace and a table (KubeDB Cassandra) |
| CockroachDB | Yes (haEnabled) | A CockroachDB cluster and a user |
| Consul | Yes | A Consul server |
| CouchDB | No | A CouchDB server with a database |
| etcd | Yes (haEnabled) | An etcd v3 cluster (KubeDB Etcd) |
| Filesystem | No | A PersistentVolumeClaim (created from volumeClaimTemplate) |
| Google Cloud Spanner | Yes (haEnabled) | A Spanner database with the Vault tables and a service account key |
| Google Cloud Storage | Yes (haEnabled) | A GCS bucket and a service account key |
| In-Memory | No | Nothing. Data is lost on restart |
| Microsoft SQL Server | No | A SQL Server (KubeDB MSSQLServer) |
| MySQL | Yes (haEnabled) | A MySQL server (KubeDB MySQL) |
| OCI Object Storage | Yes (haEnabled) | Two OCI buckets and an API signing key |
| OpenStack Swift | No | A Swift container and OpenStack credentials |
| PostgreSQL | Yes (haEnabled) | A PostgreSQL server (KubeDB Postgres) |
| Raft | Yes | A StorageClass for the integrated storage volumes |
| ZooKeeper | Yes | A ZooKeeper ensemble (KubeDB ZooKeeper) |
Things to Know
- Namespace and names. Every guide uses the
demonamespace and aVaultServernamedvault. Follow theCleanupsection of a guide before you start another one in the same namespace. - High availability. Backends without high availability have no leader election, so run them with a single replica. Backends marked
haEnabledrun in HA mode only when that field is set. Raft, Consul and ZooKeeper are always HA-capable. - Credentials. Credentials are read from a
Secretin the namespace of theVaultServer, referenced bycredentialSecretRef. For the databases that KubeDB manages,databaseRefpoints to theAppBindingof the database instead, and the operator takes the address, the credentials and the TLS settings from it. - Cleanup order. With
terminationPolicy: WipeOut, the operator connects to the storage backend to delete the data Vault stored there. Always delete theVaultServerbefore the storage backend it uses.








