: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 { DataModelingNode, DataModelingNodeType, DataModelingSyncInterval } from '~/types'
export const SYNC_INTERVAL_SECONDS: Record<DataModelingSyncInterval, number> = {
'15min': 900,
'30min': 1800,
'1hour': 3600,
'6hour': 21600,
'12hour': 43200,
'24hour': 86400,
'7day': 604800,
'30day': 2592000,
}
/** A plain view runs its query on read, so it stores nothing that can go stale. */
export const MATERIALIZING_TYPES = new Set<DataModelingNodeType>(['matview', 'endpoint'])
/**
* Age climbs to a full interval between every pair of runs, so a one-interval bar would
* flag the whole fleet just before each tick. Two means a scheduled run went missing.
*/
export const BEHIND_MULTIPLIER = 2
export interface BehindScheduleModel {
node: DataModelingNode
intervalSeconds: number
/** Seconds since the last successful run, or since creation when it has never finished one. */
ageSeconds: number
/** How far past the target the model has drifted. This is the number the table shows. */
overdueSeconds: number
}
/**
* Models whose stored data is older than the refresh target they declare. Clearing a
* target pauses a model on purpose, so a model without one is never behind.
*/
export function modelsBehindSchedule(nodes: DataModelingNode[], now: number): BehindScheduleModel[] {
const behind: BehindScheduleModel[] = []
for (const node of nodes) {
// A downstream view can make a model run more often than its own setting.
const runsAt = node.sync_interval ? (node.effective_sync_interval ?? node.sync_interval) : undefined
if (!runsAt || !MATERIALIZING_TYPES.has(node.type)) {
continue
}
const intervalSeconds = SYNC_INTERVAL_SECONDS[runsAt]
if (!intervalSeconds) {
continue
}
// A model that has never run is measured from its creation, so a new one is not
// reported before its first tick was even due.
const since = node.last_run_at ?? node.created_at
const ageSeconds = (now - new Date(since).getTime()) / 1000
if (ageSeconds <= intervalSeconds * BEHIND_MULTIPLIER) {
continue
}
behind.push({ node, intervalSeconds, ageSeconds, overdueSeconds: ageSeconds - intervalSeconds })
}
return behind.sort((a, b) => b.overdueSeconds - a.overdueSeconds)
}