Performance standby nodes

Vault supports a multi-server mode for high availability. This mode protects against outages by running multiple Vault servers. High availability mode is automatically enabled when using a data store that supports it. You can learn more about HA mode on the Concepts page.

Vault Enterprise offers additional features that allow HA nodes to service read-only requests on the local standby node. Read-only requests are requests that do not modify Vault's storage.

Server-to-Server communication

Performance Standbys require the request forwarding method described in the HA Server-to-Server docs. Unlike regular standby nodes, performance standbys will not respect the X-Vault-No-Request-Forwarding header.

A performance standby will connect to the active node over the existing request forwarding connection. If selected by the active node to be promoted to a performance standby it will be handed a newly-generated private key and certificate for use in creating a new mutually-authenticated TLS connection to the cluster port. This connection will be used to send updates from the active node to the standby.

Request forwarding

A Performance Standby will attempt to process requests that come in. If a storage write is detected the standby will forward the request over the cluster port connection to the active node. If the request is read-only the Performance Standby will handle the requests locally.

Sending requests to Performance Standbys that result in forwarded writes will be slightly slower than going directly to the active node. A client that has advanced knowledge of the behavior of the call can choose to point the request to the appropriate node.

Direct access

A Performance Standby will tag itself as such in consul if service registration is enabled. To access the set of Performance Standbys the performance-standby tag can be used. For example to send requests to only the performance standbys https://performance-standby.vault.dc1.consul could be used (host name may vary based on consul configuration).

Behind load balancers

Additionally, if you wish to point your load balancers at performance standby nodes, the sys/health endpoint can be used to determine if a node is a performance standby. See the sys/health API docs for more info.

Disabling performance standbys

To disable performance standbys the disable_performance_standby flag should be set to true in the Vault config file. This will both tell a standby not to attempt to enable performance mode and an active node to not allow any performance standby connections.

This setting should be synced across all nodes in the cluster.

Monitoring performance standbys

To verify your node is a performance standby the vault status command can be used:

$ vault status
Key                                    Value
---                                    -----
Seal Type                              shamir
Sealed                                 false
Total Shares                           1
Threshold                              1
Version                                0.11.0+ent
Cluster Name                           vault-cluster-d040e74c
Cluster ID                             9f82e03b-71fb-97a6-9c5a-46fa6715d6e4
HA Enabled                             true
HA Cluster                             https://127.0.0.1:8201
HA Mode                                standby
Active Node Address                    http://127.0.0.1:8200
Performance Standby Node               true
Performance Standby Last Remote WAL    380329