:hedgehog: PostHog is the leading platform for building self-driving products. Our developer tools – AI observability, analytics, session replay, flags, experiments, error tracking, logs, and more – capture all the context agents need to diagnose problems, uncover opportunities, and ship fixes. Steer it all from Slack, web, desktop, or the MCP.
Python ab-testingai-analyticsanalyticscdpdata-warehouseexperimentsfeature-flagsjavascriptproduct-analyticspythonreactsession-replaysurveystypescriptweb-analytics
import { Counter, Gauge, register } from 'prom-client'
import { batchAccepted, batchReleased } from './batch-metrics'
describe('batch metrics', () => {
let clock: jest.SpyInstance
async function eventsInFlight(): Promise<number> {
const gauge = register.getSingleMetric('ingestion_api_events_in_flight') as Gauge<string>
const { values } = await gauge.get()
return values[0]?.value ?? 0
}
async function eventSecondsInFlight(): Promise<number> {
const counter = register.getSingleMetric('ingestion_api_event_seconds_in_flight_total') as Counter<string>
const { values } = await counter.get()
return values[0]?.value ?? 0
}
beforeEach(() => {
clock = jest.spyOn(Date, 'now').mockReturnValue(0)
register.getSingleMetric('ingestion_api_events_in_flight')?.reset()
register.getSingleMetric('ingestion_api_event_seconds_in_flight_total')?.reset()
})
afterEach(() => {
clock.mockRestore()
})
// The processor autoscaler reads events in flight: a count keyed to the
// wrong batch, or a leaked increment, scales the fleet on phantom load.
it('sums events across concurrent batches and releases exactly the finished batch', async () => {
const first = batchAccepted(100)
const second = batchAccepted(5)
const third = batchAccepted(20)
expect(await eventsInFlight()).toBe(125)
batchReleased(second)
expect(await eventsInFlight()).toBe(120)
batchReleased(third)
expect(await eventsInFlight()).toBe(100)
batchReleased(first)
expect(await eventsInFlight()).toBe(0)
})
// rate() over the counter must recover the time-weighted mean events in
// flight, so each batch has to be credited its own events * duration.
it('credits each batch its own event-seconds', async () => {
const long = batchAccepted(100)
clock.mockReturnValue(1_000)
const short = batchAccepted(4)
// short: accepted at 1s, released at 3s -> 4 * 2 = 8
clock.mockReturnValue(3_000)
batchReleased(short)
expect(await eventSecondsInFlight()).toBe(8)
// long: accepted at 0s, released at 4s -> 100 * 4 = 400
clock.mockReturnValue(4_000)
batchReleased(long)
expect(await eventSecondsInFlight()).toBe(408)
})
})