Merge pull request #3866 from Infisical/feat/posthogEventBatch

feat(telemetry): Add aggregated events and groups to posthog
This commit is contained in:
carlosmonastyrski
2025-06-27 14:42:55 -03:00
committed by GitHub
21 changed files with 404 additions and 17 deletions
@@ -7,13 +7,18 @@ import { QueueJobs, QueueName, TQueueServiceFactory } from "@app/queue";
import { getServerCfg } from "../super-admin/super-admin-service";
import { TTelemetryDALFactory } from "./telemetry-dal";
import { TELEMETRY_SECRET_OPERATIONS_KEY, TELEMETRY_SECRET_PROCESSED_KEY } from "./telemetry-service";
import {
TELEMETRY_SECRET_OPERATIONS_KEY,
TELEMETRY_SECRET_PROCESSED_KEY,
TTelemetryServiceFactory
} from "./telemetry-service";
import { PostHogEventTypes } from "./telemetry-types";
type TTelemetryQueueServiceFactoryDep = {
queueService: TQueueServiceFactory;
keyStore: Pick<TKeyStoreFactory, "getItem" | "deleteItem">;
telemetryDAL: TTelemetryDALFactory;
telemetryService: TTelemetryServiceFactory;
};
export type TTelemetryQueueServiceFactory = ReturnType<typeof telemetryQueueServiceFactory>;
@@ -21,7 +26,8 @@ export type TTelemetryQueueServiceFactory = ReturnType<typeof telemetryQueueServ
export const telemetryQueueServiceFactory = ({
queueService,
keyStore,
telemetryDAL
telemetryDAL,
telemetryService
}: TTelemetryQueueServiceFactoryDep) => {
const appCfg = getConfig();
const postHog =
@@ -48,6 +54,10 @@ export const telemetryQueueServiceFactory = ({
await keyStore.deleteItem(TELEMETRY_SECRET_OPERATIONS_KEY);
});
queueService.start(QueueName.TelemetryAggregatedEvents, async () => {
await telemetryService.processAggregatedEvents();
});
// every day at midnight a telemetry job executes on self-hosted instances
// this sends some telemetry information like instance id secrets operated etc
const startTelemetryCheck = async () => {
@@ -60,11 +70,26 @@ export const telemetryQueueServiceFactory = ({
{ pattern: "0 0 * * *", utc: true },
QueueName.TelemetryInstanceStats // just a job id
);
// clear previous aggregated events job
await queueService.stopRepeatableJob(
QueueName.TelemetryAggregatedEvents,
QueueJobs.TelemetryAggregatedEvents,
{ pattern: "*/5 * * * *", utc: true },
QueueName.TelemetryAggregatedEvents // just a job id
);
if (postHog) {
await queueService.queue(QueueName.TelemetryInstanceStats, QueueJobs.TelemetryInstanceStats, undefined, {
jobId: QueueName.TelemetryInstanceStats,
repeat: { pattern: "0 0 * * *", utc: true }
});
// Start aggregated events job (runs every five minutes)
await queueService.queue(QueueName.TelemetryAggregatedEvents, QueueJobs.TelemetryAggregatedEvents, undefined, {
jobId: QueueName.TelemetryAggregatedEvents,
repeat: { pattern: "*/5 * * * *", utc: true }
});
}
};
@@ -72,6 +97,10 @@ export const telemetryQueueServiceFactory = ({
logger.error(err?.failedReason, `${QueueName.TelemetryInstanceStats}: failed`);
});
queueService.listen(QueueName.TelemetryAggregatedEvents, "failed", (err) => {
logger.error(err?.failedReason, `${QueueName.TelemetryAggregatedEvents}: failed`);
});
return {
startTelemetryCheck
};
@@ -1,3 +1,4 @@
import { createHash, randomUUID } from "crypto";
import { PostHog } from "posthog-node";
import { TLicenseServiceFactory } from "@app/ee/services/license/license-service";
@@ -12,12 +13,49 @@ import { PostHogEventTypes, TPostHogEvent, TSecretModifiedEvent } from "./teleme
export const TELEMETRY_SECRET_PROCESSED_KEY = "telemetry-secret-processed";
export const TELEMETRY_SECRET_OPERATIONS_KEY = "telemetry-secret-operations";
export const POSTHOG_AGGREGATED_EVENTS = [PostHogEventTypes.SecretPulled];
const TELEMETRY_AGGREGATED_KEY_EXP = 900; // 15mins
// Bucket configuration
const TELEMETRY_BUCKET_COUNT = 30;
const TELEMETRY_BUCKET_NAMES = Array.from(
{ length: TELEMETRY_BUCKET_COUNT },
(_, i) => `bucket-${i.toString().padStart(2, "0")}`
);
type AggregatedEventData = Record<string, unknown>;
type SingleEventData = {
distinctId: string;
event: string;
properties: unknown;
organizationId: string;
};
export type TTelemetryServiceFactory = ReturnType<typeof telemetryServiceFactory>;
export type TTelemetryServiceFactoryDep = {
keyStore: Pick<TKeyStoreFactory, "getItem" | "incrementBy">;
keyStore: Pick<
TKeyStoreFactory,
"incrementBy" | "deleteItemsByKeyIn" | "setItemWithExpiry" | "getKeysByPattern" | "getItems"
>;
licenseService: Pick<TLicenseServiceFactory, "getInstanceType">;
};
const getBucketForDistinctId = (distinctId: string): string => {
// Use SHA-256 hash for consistent distribution
const hash = createHash("sha256").update(distinctId).digest("hex");
// Take first 8 characters and convert to number for better distribution
const hashNumber = parseInt(hash.substring(0, 8), 16);
const bucketIndex = hashNumber % TELEMETRY_BUCKET_COUNT;
return TELEMETRY_BUCKET_NAMES[bucketIndex];
};
export const createTelemetryEventKey = (event: string, distinctId: string): string => {
const bucketId = getBucketForDistinctId(distinctId);
return `telemetry-event-${event}-${bucketId}-${distinctId}-${randomUUID()}`;
};
export const telemetryServiceFactory = ({ keyStore, licenseService }: TTelemetryServiceFactoryDep) => {
const appCfg = getConfig();
@@ -64,11 +102,33 @@ To opt into telemetry, you can set "TELEMETRY_ENABLED=true" within the environme
const instanceType = licenseService.getInstanceType();
// capture posthog only when its cloud or signup event happens in self-hosted
if (instanceType === InstanceType.Cloud || event.event === PostHogEventTypes.UserSignedUp) {
postHog.capture({
event: event.event,
distinctId: event.distinctId,
properties: event.properties
});
if (event.organizationId) {
try {
postHog.groupIdentify({ groupType: "organization", groupKey: event.organizationId });
} catch (error) {
logger.error(error, "Failed to identify PostHog organization");
}
}
if (POSTHOG_AGGREGATED_EVENTS.includes(event.event)) {
const eventKey = createTelemetryEventKey(event.event, event.distinctId);
await keyStore.setItemWithExpiry(
eventKey,
TELEMETRY_AGGREGATED_KEY_EXP,
JSON.stringify({
distinctId: event.distinctId,
event: event.event,
properties: event.properties,
organizationId: event.organizationId
})
);
} else {
postHog.capture({
event: event.event,
distinctId: event.distinctId,
properties: event.properties,
...(event.organizationId ? { groups: { organization: event.organizationId } } : {})
});
}
return;
}
@@ -89,6 +149,160 @@ To opt into telemetry, you can set "TELEMETRY_ENABLED=true" within the environme
}
};
const aggregateGroupProperties = (events: SingleEventData[]): AggregatedEventData => {
const aggregatedData: AggregatedEventData = {};
// Set the total count
aggregatedData.count = events.length;
events.forEach((event) => {
if (!event.properties) return;
Object.entries(event.properties as Record<string, unknown>).forEach(([key, value]: [string, unknown]) => {
if (Array.isArray(value)) {
// For arrays, count occurrences of each item
const existingCounts =
aggregatedData[key] &&
typeof aggregatedData[key] === "object" &&
aggregatedData[key]?.constructor === Object
? (aggregatedData[key] as Record<string, number>)
: {};
value.forEach((item) => {
const itemKey = typeof item === "object" ? JSON.stringify(item) : String(item);
existingCounts[itemKey] = (existingCounts[itemKey] || 0) + 1;
});
aggregatedData[key] = existingCounts;
} else if (typeof value === "object" && value?.constructor === Object) {
// For objects, count occurrences of each field value
const existingCounts =
aggregatedData[key] &&
typeof aggregatedData[key] === "object" &&
aggregatedData[key]?.constructor === Object
? (aggregatedData[key] as Record<string, number>)
: {};
if (value) {
Object.values(value).forEach((fieldValue) => {
const valueKey = typeof fieldValue === "object" ? JSON.stringify(fieldValue) : String(fieldValue);
existingCounts[valueKey] = (existingCounts[valueKey] || 0) + 1;
});
}
aggregatedData[key] = existingCounts;
} else if (typeof value === "number") {
// For numbers, add to existing sum
aggregatedData[key] = ((aggregatedData[key] as number) || 0) + value;
} else if (value !== undefined && value !== null) {
// For other types (strings, booleans, etc.), count occurrences
const stringValue = String(value);
const existingValue = aggregatedData[key];
if (!existingValue) {
aggregatedData[key] = { [stringValue]: 1 };
} else if (existingValue && typeof existingValue === "object" && existingValue.constructor === Object) {
const countObject = existingValue as Record<string, number>;
countObject[stringValue] = (countObject[stringValue] || 0) + 1;
} else {
const oldValue = String(existingValue);
aggregatedData[key] = {
[oldValue]: 1,
[stringValue]: 1
};
}
}
});
});
return aggregatedData;
};
const processBucketEvents = async (eventType: string, bucketId: string) => {
if (!postHog) return 0;
try {
const bucketPattern = `telemetry-event-${eventType}-${bucketId}-*`;
const bucketKeys = await keyStore.getKeysByPattern(bucketPattern);
if (bucketKeys.length === 0) return 0;
const bucketEvents = await keyStore.getItems(bucketKeys);
let bucketEventsParsed: SingleEventData[] = [];
try {
bucketEventsParsed = bucketEvents
.filter((event) => event !== null)
.map((event) => JSON.parse(event as string) as SingleEventData);
} catch (error) {
logger.error(error, `Failed to parse bucket events for ${eventType} in ${bucketId}`);
return 0;
}
const eventsGrouped = new Map<string, SingleEventData[]>();
bucketEventsParsed.forEach((event) => {
const key = JSON.stringify({ id: event.distinctId, org: event.organizationId });
if (!eventsGrouped.has(key)) {
eventsGrouped.set(key, []);
}
eventsGrouped.get(key)!.push(event);
});
if (eventsGrouped.size === 0) return 0;
for (const [eventsKey, events] of eventsGrouped) {
const key = JSON.parse(eventsKey) as { id: string; org?: string };
if (key.org) {
try {
postHog.groupIdentify({ groupType: "organization", groupKey: key.org });
} catch (error) {
logger.error(error, "Failed to identify PostHog organization");
}
}
const properties = aggregateGroupProperties(events);
postHog.capture({
event: `${eventType} aggregated`,
distinctId: key.id,
properties,
...(key.org ? { groups: { organization: key.org } } : {})
});
}
// Clean up processed data for this bucket
await keyStore.deleteItemsByKeyIn(bucketKeys);
logger.info(`Processed ${bucketEventsParsed.length} events from bucket ${bucketId} for ${eventType}`);
return bucketEventsParsed.length;
} catch (error) {
logger.error(error, `Failed to process bucket ${bucketId} for ${eventType}`);
return 0;
}
};
const processAggregatedEvents = async () => {
if (!postHog) return;
for (const eventType of POSTHOG_AGGREGATED_EVENTS) {
let totalProcessed = 0;
logger.info(`Starting bucket processing for ${eventType}`);
// Process each bucket sequentially to control memory usage
for (const bucketId of TELEMETRY_BUCKET_NAMES) {
try {
// eslint-disable-next-line no-await-in-loop
const processed = await processBucketEvents(eventType, bucketId);
totalProcessed += processed;
} catch (error) {
logger.error(error, `Failed to process bucket ${bucketId} for ${eventType}`);
}
}
logger.info(`Completed processing ${totalProcessed} total events for ${eventType}`);
}
};
const flushAll = async () => {
if (postHog) {
await postHog.shutdownAsync();
@@ -98,6 +312,8 @@ To opt into telemetry, you can set "TELEMETRY_ENABLED=true" within the environme
return {
sendLoopsEvent,
sendPostHogEvents,
flushAll
processAggregatedEvents,
flushAll,
getBucketForDistinctId
};
};
@@ -1,3 +1,13 @@
import {
IdentityActor,
KmipClientActor,
PlatformActor,
ScimClientActor,
ServiceActor,
UnknownUserActor,
UserActor
} from "@app/ee/services/audit-log/audit-log-types";
export enum PostHogEventTypes {
SecretPush = "secrets pushed",
SecretPulled = "secrets pulled",
@@ -40,6 +50,14 @@ export type TSecretModifiedEvent = {
secretPath: string;
channel?: string;
userAgent?: string;
actor?:
| UserActor
| IdentityActor
| ServiceActor
| ScimClientActor
| PlatformActor
| UnknownUserActor
| KmipClientActor;
};
};
@@ -214,7 +232,7 @@ export type TInvalidateCacheEvent = {
};
};
export type TPostHogEvent = { distinctId: string } & (
export type TPostHogEvent = { distinctId: string; organizationId?: string } & (
| TSecretModifiedEvent
| TAdminInitEvent
| TUserSignedUpEvent