BlitzQ
API Reference

metrics

API reference for metrics.

Low-overhead in-process metrics.

Counters and a fixed-bucket duration histogram are kept per queue (never per task id, so label cardinality stays bounded by the number of queues). Updating a metric is a dict lookup and an integer add. Workers publish periodic snapshots to the broker for blitzq queue stats; an optional Prometheus exporter is available with pip install "blitzq[monitoring]".

class Metrics

__init__() -> None

inc(queue: str, event: str, n: int = 1) -> None

observe_duration(queue: str, seconds: float) -> None

observe_latency(queue: str, seconds: float) -> None

Enqueue-to-start latency (wall clock; subject to cross-host clock skew).

snapshot() -> dict[str, Any]

start_prometheus_exporter(metrics: Metrics, port: int, addr: str = '0.0.0.0') -> None

Serve metrics on http://addr:port/metrics (requires prometheus-client).

worker

API reference for worker.

On this page

class Metrics__init__() -> Noneinc(queue: str, event: str, n: int = 1) -> Noneobserve_duration(queue: str, seconds: float) -> Noneobserve_latency(queue: str, seconds: float) -> Nonesnapshot() -> dict[str, Any]start_prometheus_exporter(metrics: Metrics, port: int, addr: str = '0.0.0.0') -> None