Struct ReadWriteClient 

pub struct ReadWriteClient { /* private fields */ }

Read/write-split cluster client. Owns one RespClient to the primary node + one per replica node. Round-robins reads across the replica fleet (fallback to primary on empty fleet or consistent = true).

Implementations

impl ReadWriteClient

pub fn connect(primary: (&str, u16), replicas: &[(&str, u16)]) -> Result<Self>

Open one connection to the primary and one per replica.

primary is the address of the kevy node running with [replication] role = "primary" (or standalone — both behave the same from the client’s perspective). replicas lists the addresses of kevy nodes running with [replication] role = "replica"; v1.18 assumes they all share the keyspace of the primary (operator-enforced — kevy has no automatic discovery in v1.18).

pub fn replica_count(&self) -> usize

Number of replica connections.

pub fn request_write(&mut self, args: &[Vec<u8>]) -> Result<Reply>

Route a write command (or any command that must hit the primary) to the primary connection. If the server replies -MISDIRECTED writer is <host:port> (Phase 3 / v1.21 scoped multi-writer), the client transparently opens a connection to the named writer (caching it), caches the key→writer mapping for follow-up writes on the same key, and retries once. A second -MISDIRECTED from the retry surfaces as an error.

pub fn request_read( &mut self, args: &[Vec<u8>], consistent: bool, ) -> Result<Reply>

Route a read command to a replica (round-robin); fallback to the primary when no replica is configured or when consistent is true.

pub fn request(&mut self, args: &[Vec<u8>]) -> Result<Reply>

Auto-routed command. Classifies args[0] via is_write_verb and dispatches to either Self::request_write or Self::request_read (consistent = false). Convenience for callers that don’t want to make the read/write decision explicit.