Struct FailoverWatchdog 

pub struct FailoverWatchdog<'a> {
    pub table: &'a HeartbeatTable,
    pub grace_epochs: u64,
}

Watchdog scanner; one per cooperating cluster.

Fields

table: &'a HeartbeatTable

grace_epochs: u64

Implementations

impl<'a> FailoverWatchdog<'a>

pub fn new(table: &'a HeartbeatTable) -> Self

pub fn with_grace(table: &'a HeartbeatTable, grace_epochs: u64) -> Self

pub fn scan(&self) -> ReclaimReport

Advance global epoch and scan every slot. Returns a report of slots whose last beat is more than grace_epochs behind the new global epoch.

Per-scan observation is pushed to the underlying HeartbeatTable’s sidecar ring rather than a separate watchdog-owned ring: the watchdog borrows the table with lifetime 'a, which is incompatible with the AdaptiveInstance: 'static trait bound. Routing the scan observation through the table preserves visibility to any policy attached at that level.

pub fn iter_in_flight_bits(bitmap: u64) -> impl Iterator<Item = u8>

Iterate the set bits in bitmap, returning each bit’s position. Used by callers reclaiming an in_flight_bitmap.

pub fn clear_dead_bitmap(&self, slot_idx: usize)

Clear the dead process’s bitmap so subsequent scans don’t re-report it. Typically called by the caller after they have reassigned the work.