Skip to main content

WIDE_LATENCY_BOUNDS_US

Constant WIDE_LATENCY_BOUNDS_US 

Source
pub const WIDE_LATENCY_BOUNDS_US: &[u64];
Expand description

Wide latency bucket upper bounds in microseconds: 1ms … 300s.

Use for histograms that measure spans dominated by network I/O (upstream LLM streaming, reconciler recovery scans, finalisation under load). Provider p99 for GPT-4o / Claude regularly sits in the 15–90 s band; the top bound of 300 s covers pathological long contexts without saturating for realistic operator SLA use.