Struct PromClient 

pub struct PromClient { /* private fields */ }

Implementations

impl PromClient

pub fn new(base_url: impl Into<String>, request_timeout: Time) -> Result<Self>

base_url is the Prometheus root, for example http://prom.monitoring.svc:9090. A trailing slash is not needed.

Errors

Returns an error when input data is invalid, required I/O fails, or the destination rejects the generated report or audit event.

pub async fn query_scalar_sum(&self, query: &str) -> Result<Option<f64>>

Runs a PromQL instant query. Returns the sum of the first scalar value across all returned series. Returns None when the result vector is empty.

Errors

Returns an error when input data is invalid, required I/O fails, or the destination rejects the generated report or audit event.

pub async fn query_scalar_sum_at( &self, query: &str, at_unix_s: Option<f64>, ) -> Result<Option<f64>>

Self::query_scalar_sum, evaluated at an explicit instant.

at_unix_s is the Prometheus time parameter. A range selector inside the query then looks back from that instant and not from now, which is what lets a caller name the window it measured rather than the window that happens to end when the client got round to asking.

Errors

Returns an error when input data is invalid, required I/O fails, or the destination rejects the generated report or audit event.

pub async fn query_range_sum( &self, query: &str, start_s: f64, end_s: f64, step_s: u64, ) -> Result<Vec<PromPoint>>

Runs a PromQL range query and sums across all returned series per timestamp. Returns PromPoints on the step grid.

Errors

Returns an error when input data is invalid, required I/O fails, or the destination rejects the generated report or audit event.

pub async fn capture_resource_series( &self, stack: Stack, namespace: &str, start_s: f64, end_s: f64, step_s: u64, ) -> Result<Vec<BrokerSample>>

Captures a broker CPU and memory time series over [start_s, end_s], so a graph can show the values over the test. CPU is a 1-minute rate() in cores, and memory is the summed working set. Both align onto a single step_s grid. t_offset_ms is relative to start_s.

Errors

Returns an error when input data is invalid, required I/O fails, or the destination rejects the generated report or audit event.

pub async fn capture_resource( &self, stack: Stack, namespace: &str, window: Time, measured_until_unix_s: f64, msgs_produced: MessageCount, ) -> Result<Resource>

Captures broker resource usage for the given stack over a window that ends now. Make window a little larger than the measured scenario duration, so the rate() window does not tail off.

Errors

Returns an error when input data is invalid, required I/O fails, or the destination rejects the generated report or audit event. measured_until_unix_s is the instant the measurement window closed. The queries are evaluated there rather than at call time, so the window-long lookback covers the window that was measured. Without it, the time the driver spends joining its tasks and draining its producers shifts the lookback forward by that much: the start of the measurement falls out of the window and the drain falls into it, while the efficiency figure still divides the measurement-only message count by that shifted CPU interval.