Vault Deployment Architecture · Topic 1 of 3
Cluster Strategy and Deployment Models
Vault Associate 003: the recommended cluster layout, five nodes across three zones, and how self-managed Vault differs from HCP Vault Dedicated.
- 4 min
- A cluster has one active node and standby nodes, with a load balancer in front.
- The recommended single cluster is five nodes in three availability zones, tolerating two node failures.
- Use an odd number of nodes, and do not stretch one cluster across regions.
What it is
A Vault cluster is a group of Vault servers that share one storage backend and act as a single service. Your cluster strategy is how many nodes you run, where you place them and how clients reach them.
There are two deployment models. In a self-managed cluster you install, configure and operate Vault yourself. In a HashiCorp-managed cluster, HashiCorp runs the infrastructure for you as HCP Vault Dedicated.
How it works
Active and standby nodes
When a cluster uses a storage backend that supports high availability, one node becomes the active node and the others are standbys. The active node serves requests. A standby either forwards a request to the active node or redirects the client to it.
Clients therefore do not care which node they hit, as long as the node is unsealed. A load balancer in front of the cluster spreads the connections.
The recommended layout
The reference architecture for integrated storage recommends a single cluster of five nodes spread across three availability zones. That layout tolerates losing two nodes or one whole availability zone.
- Use an odd number of nodes so votes cannot tie.
- Keep latency between zones low. The reference architecture gives under 8 ms.
- Point the load balancer health checks at the
/v1/sys/healthendpoint. - A single cluster is not spread across regions because of network latency.
To add resilience across regions or clusters, you add more clusters and link them with Vault Enterprise replication, which the third page of this chapter covers.
Self-managed or HashiCorp-managed
HCP Vault Dedicated is a managed Vault Enterprise service. It runs the same Vault as self-hosted Vault Enterprise, so the CLI, API and UI work the same way.
| Area | Self-managed | HCP Vault Dedicated |
|---|---|---|
| Servers and infrastructure | You provision and run them | HashiCorp |
| Storage, backups and snapshots | You | HashiCorp |
| Upgrades | You plan and apply them | HashiCorp |
| High availability and recovery | You design it | HashiCorp |
| Policies, auth methods, secrets engines | You | You |
| Application integration, capacity planning | You | You |
| Edition | Community or Enterprise | Enterprise |
| Seal | Shamir by default | Auto unseal set up for you, with a unique KMS key per cluster |
| Cluster size | No built-in scaling feature | Scale from the HCP portal or Terraform |
In HCP Vault Dedicated the top-level namespace is admin instead of root. Namespaces, replication and performance standbys are Enterprise features that Community Edition lacks. HCP Vault Dedicated includes most Enterprise features, but replication needs its Standard tier, and its Development tier is a single node for non-production use.
How the exam asks about this
- "A team has no staff to patch servers, take backups or manage cluster health. Which model fits?" HCP Vault Dedicated, because HashiCorp runs the infrastructure.
- "Which layout does the reference architecture recommend for one cluster?" Five nodes across three availability zones.
- "Who writes the policies and enables the secrets engines in an HCP Vault Dedicated cluster?" The customer. HashiCorp manages the platform, not your Vault configuration.
What to remember for the exam
- A cluster has one active node and standby nodes, with a load balancer in front.
- The recommended single cluster is five nodes in three availability zones, tolerating two node failures.
- Use an odd number of nodes, and do not stretch one cluster across regions.
- HCP Vault Dedicated is Vault Enterprise run by HashiCorp, with the same CLI, API and UI.
- HashiCorp manages upgrades, backups, storage operations and availability. You still manage policies, auth methods and secrets engines.
- Replication across clusters is an Enterprise feature, also available on the HCP Vault Dedicated Standard tier.
HashiCorp-managed does not mean hands off. Authentication, access policies and secrets engines are still the customer's job.
Common mistakes
Thinking more nodes always means more throughput. Standby nodes in a cluster exist for fault tolerance. Scaling reads across sites is the job of performance replication, not extra standbys, and Community Edition standbys do not serve requests themselves.
Choosing an even node count. Four nodes tolerate no more failures than three, and six no more than five. An odd count is the recommended design.
Assuming HCP Vault Dedicated is a different product. It runs the same Vault Enterprise software. The difference is who operates it, not what the CLI or API can do.
Check yourself
How many nodes and zones does the reference architecture recommend for one cluster?
Five nodes across three availability zones. That tolerates the loss of two nodes or one availability zone.
Which tasks stay with the customer on HCP Vault Dedicated?
Configuring Vault itself: auth methods, policies, secrets engines and application integration. HashiCorp runs the platform, backups and upgrades.
Can one Vault cluster be stretched across regions?
No, the reference architecture avoids it because of latency. Use replication between separate clusters instead.
Topic 1 of 3 in Vault Deployment Architecture
Test yourself on Vault Deployment Architecture
The full HashiCorp Vault Associate 003 question bank has 600 questions, with timed practice exams and an explanation for every answer.
This study guide is independent and is not affiliated with, endorsed by, or sponsored by HashiCorp. Vendor names, certification names and exam codes are trademarks of their respective owners and are used for identification only.