Struct RpcHealthStore

pub struct RpcHealthStore { /* private fields */ }

Shared in-memory store for RPC health metadata.

This store is shared across all relayers, so health state for a given RPC URL is consistent across all relayers using that URL.

Implementations

impl RpcHealthStore

pub fn instance() -> &'static RpcHealthStore

Gets the singleton instance of the health store.

pub fn get_metadata(&self, url: &str) -> RpcConfigMetadata

Gets metadata for a given RPC URL.

Returns default (empty) metadata if the URL is not in the store.

Arguments

• url - The RPC endpoint URL

Returns

• RpcConfigMetadata - The metadata for the URL, or default if not found

pub fn update_metadata(&self, url: &str, metadata: RpcConfigMetadata)

Updates metadata for a given RPC URL.

Arguments

• url - The RPC endpoint URL
• metadata - The metadata to store

pub fn mark_failed( &self, url: &str, threshold: u32, pause_duration: Duration, failure_expiration: Duration, )

Marks an RPC endpoint as failed, adding a failure timestamp. If the number of recent failures (within expiration window) reaches the threshold, pauses the endpoint. Stale failures (older than failure_expiration) are automatically removed.

Arguments

• url - The RPC endpoint URL
• threshold - The number of failures before pausing
• pause_duration - The duration to pause for
• failure_expiration - Duration after which failures are considered stale and removed

pub fn reset_failures(&self, url: &str)

Resets the failure count and unpauses the endpoint.

Arguments

• url - The RPC endpoint URL

pub fn reset_failures_if_exists(&self, url: &str) -> bool

Resets the failure count only if the provider has failures recorded.

Arguments

• url - The RPC endpoint URL

Returns

• bool - True if failures were reset, false if no failures existed

pub fn is_paused( &self, url: &str, threshold: u32, failure_expiration: Duration, ) -> bool

Checks if an RPC endpoint is currently paused.

An endpoint is considered paused if:

• It has reached the failure threshold (within expiration window) AND
• It has a paused_until timestamp that is in the future

Stale failures (older than failure_expiration) are automatically removed to allow the provider to be retried.

This method uses a read-lock-first pattern for better concurrency:

• First acquires a read lock to check if modification is needed
• Only upgrades to write lock if cleanup is required

Arguments

• url - The RPC endpoint URL
• threshold - The failure threshold to check against
• failure_expiration - Duration after which failures are considered stale and removed

Returns

• bool - True if the endpoint is paused, false otherwise