Performance standby

Performance standby lets the non-active HA nodes inside a single cluster serve read requests locally instead of forwarding every request to the active node. This spreads read load across all cluster nodes and reduces latency, while writes and other operations that change state are still forwarded to the active node.

Unlike Performance replication and Disaster recovery, which operate across separate clusters, performance standby works within one cluster and requires no cross-cluster setup.

How it works

• In an HA cluster, one node is active and the others are standby nodes.
• In Stronghold EE, each standby node becomes a performance standby: it handles read requests against its local state.
• Requests that modify state (writes, deletes) are forwarded to the active node and applied there.

Performance standby is enabled by default in EE builds. No additional configuration is required beyond running an HA cluster on integrated Raft storage.

Verify performance standby

Send a read request directly to a standby node with the X-Vault-No-Request-Forwarding header. A performance standby still returns 200 for a read because it serves it locally; a request that requires the active node returns a non-2xx status when forwarding is suppressed.

curl \
  --header "X-Vault-Token: ${VAULT_TOKEN}" \
  --header "X-Vault-No-Request-Forwarding: true" \
  "${STANDBY_ADDR}/v1/secret/data/hello"

You can also check the node role in the health and status output — a performance standby reports itself as a standby that serves reads.

Disable performance standby

To turn off performance standby without affecting replication, set the following parameter in the server configuration:

disable_performance_standby = true

With this parameter, standby nodes forward all requests to the active node, as they would without the Enterprise standby capability.