refactor(server): standarderlize metrics and trace with OTEL (#5054)

you can now export span to Zipkin and metrics to Prometheus when developing locally
follow the docs of OTEL: https://opentelemetry.io/docs/instrumentation/js/exporters/

<img width="2357" alt="image" src="https://github.com/toeverything/AFFiNE/assets/8281226/ec615e1f-3e91-43f7-9111-d7d2629e9679">
This commit is contained in:
liuyi
2023-11-24 15:19:22 +00:00
parent cf65a5cd93
commit 91efca107a
23 changed files with 298 additions and 360 deletions
@@ -1,18 +0,0 @@
import { Controller, Get, Res } from '@nestjs/common';
import type { Response } from 'express';
import { register } from 'prom-client';
import { PrismaService } from '../prisma';
@Controller()
export class MetricsController {
constructor(private readonly prisma: PrismaService) {}
@Get('/metrics')
async index(@Res() res: Response): Promise<void> {
res.header('Content-Type', register.contentType);
const prismaMetrics = await this.prisma.$metrics.prometheus();
const appMetrics = await register.metrics();
res.send(appMetrics + prismaMetrics);
}
}
+3 -13
View File
@@ -1,13 +1,3 @@
import { Global, Module } from '@nestjs/common';
import { MetricsController } from '../metrics/controller';
import { Metrics } from './metrics';
@Global()
@Module({
providers: [Metrics],
exports: [Metrics],
controllers: [MetricsController],
})
export class MetricsModule {}
export { Metrics };
export * from './metrics';
export { start } from './opentelemetry';
export * from './utils';
+69 -24
View File
@@ -1,31 +1,76 @@
import { Injectable, OnModuleDestroy } from '@nestjs/common';
import { register } from 'prom-client';
import opentelemetry, { Attributes, Observable } from '@opentelemetry/api';
import { metricsCreator } from './utils';
interface AsyncMetric {
ob: Observable;
get value(): any;
get attrs(): Attributes | undefined;
}
@Injectable()
export class Metrics implements OnModuleDestroy {
onModuleDestroy(): void {
register.clear();
let _metrics: ReturnType<typeof createBusinessMetrics> | undefined = undefined;
export function getMeter(name = 'business') {
return opentelemetry.metrics.getMeter(name);
}
function createBusinessMetrics() {
const meter = getMeter();
const asyncMetrics: AsyncMetric[] = [];
function createGauge(name: string) {
let value: any;
let attrs: Attributes | undefined;
const ob = meter.createObservableGauge(name);
asyncMetrics.push({
ob,
get value() {
return value;
},
get attrs() {
return attrs;
},
});
return (newValue: any, newAttrs?: Attributes) => {
value = newValue;
attrs = newAttrs;
};
}
socketIOEventCounter = metricsCreator.counter('socket_io_counter', ['event']);
socketIOEventTimer = metricsCreator.timer('socket_io_timer', ['event']);
socketIOConnectionGauge = metricsCreator.gauge(
'socket_io_connection_counter'
const metrics = {
socketIOConnectionGauge: createGauge('socket_io_connection'),
gqlRequest: meter.createCounter('gql_request'),
gqlError: meter.createCounter('gql_error'),
gqlTimer: meter.createHistogram('gql_timer'),
jwstCodecMerge: meter.createCounter('jwst_codec_merge'),
jwstCodecDidnotMatch: meter.createCounter('jwst_codec_didnot_match'),
jwstCodecFail: meter.createCounter('jwst_codec_fail'),
authCounter: meter.createCounter('auth'),
authFailCounter: meter.createCounter('auth_fail'),
docHistoryCounter: meter.createCounter('doc_history_created'),
docRecoverCounter: meter.createCounter('doc_history_recovered'),
};
meter.addBatchObservableCallback(
result => {
asyncMetrics.forEach(metric => {
result.observe(metric.ob, metric.value, metric.attrs);
});
},
asyncMetrics.map(({ ob }) => ob)
);
gqlRequest = metricsCreator.counter('gql_request', ['operation']);
gqlError = metricsCreator.counter('gql_error', ['operation']);
gqlTimer = metricsCreator.timer('gql_timer', ['operation']);
jwstCodecMerge = metricsCreator.counter('jwst_codec_merge');
jwstCodecDidnotMatch = metricsCreator.counter('jwst_codec_didnot_match');
jwstCodecFail = metricsCreator.counter('jwst_codec_fail');
authCounter = metricsCreator.counter('auth');
authFailCounter = metricsCreator.counter('auth_fail', ['reason']);
docHistoryCounter = metricsCreator.counter('doc_history_created');
docRecoverCounter = metricsCreator.counter('doc_history_recovered');
return metrics;
}
export function registerBusinessMetrics() {
if (!_metrics) {
_metrics = createBusinessMetrics();
}
return _metrics;
}
export const metrics = registerBusinessMetrics;
@@ -0,0 +1,127 @@
import { MetricExporter } from '@google-cloud/opentelemetry-cloud-monitoring-exporter';
import { TraceExporter } from '@google-cloud/opentelemetry-cloud-trace-exporter';
import {
CompositePropagator,
W3CBaggagePropagator,
W3CTraceContextPropagator,
} from '@opentelemetry/core';
import { PrometheusExporter } from '@opentelemetry/exporter-prometheus';
import { ZipkinExporter } from '@opentelemetry/exporter-zipkin';
import { HostMetrics } from '@opentelemetry/host-metrics';
import { Instrumentation } from '@opentelemetry/instrumentation';
import { GraphQLInstrumentation } from '@opentelemetry/instrumentation-graphql';
import { HttpInstrumentation } from '@opentelemetry/instrumentation-http';
import { IORedisInstrumentation } from '@opentelemetry/instrumentation-ioredis';
import { NestInstrumentation } from '@opentelemetry/instrumentation-nestjs-core';
import { SocketIoInstrumentation } from '@opentelemetry/instrumentation-socket.io';
import {
ConsoleMetricExporter,
MetricReader,
PeriodicExportingMetricReader,
} from '@opentelemetry/sdk-metrics';
import { NodeSDK } from '@opentelemetry/sdk-node';
import {
BatchSpanProcessor,
ConsoleSpanExporter,
SpanExporter,
} from '@opentelemetry/sdk-trace-node';
import { PrismaInstrumentation } from '@prisma/instrumentation';
import { registerBusinessMetrics } from './metrics';
abstract class OpentelemetryFactor {
abstract getMetricReader(): MetricReader;
abstract getSpanExporter(): SpanExporter;
getInstractions(): Instrumentation[] {
return [
new NestInstrumentation(),
new IORedisInstrumentation(),
new SocketIoInstrumentation({ traceReserved: true }),
new GraphQLInstrumentation({ mergeItems: true }),
new HttpInstrumentation(),
new PrismaInstrumentation(),
];
}
create() {
const traceExporter = this.getSpanExporter();
return new NodeSDK({
traceExporter,
metricReader: this.getMetricReader(),
spanProcessor: new BatchSpanProcessor(traceExporter),
textMapPropagator: new CompositePropagator({
propagators: [
new W3CBaggagePropagator(),
new W3CTraceContextPropagator(),
],
}),
instrumentations: this.getInstractions(),
serviceName: 'affine-cloud',
});
}
}
class GCloudOpentelemetryFactor extends OpentelemetryFactor {
override getMetricReader(): MetricReader {
return new PeriodicExportingMetricReader({
exportIntervalMillis: 30000,
exportTimeoutMillis: 60000,
exporter: new MetricExporter(),
});
}
override getSpanExporter(): SpanExporter {
return new TraceExporter();
}
}
class LocalOpentelemetryFactor extends OpentelemetryFactor {
override getMetricReader(): MetricReader {
return new PrometheusExporter();
}
override getSpanExporter(): SpanExporter {
return new ZipkinExporter();
}
}
class DebugOpentelemetryFactor extends OpentelemetryFactor {
override getMetricReader(): MetricReader {
return new PeriodicExportingMetricReader({
exporter: new ConsoleMetricExporter(),
});
}
override getSpanExporter(): SpanExporter {
return new ConsoleSpanExporter();
}
}
function createSDK() {
let factor: OpentelemetryFactor | null = null;
if (process.env.NODE_ENV === 'production') {
factor = new GCloudOpentelemetryFactor();
} else if (process.env.DEBUG_METRICS) {
factor = new DebugOpentelemetryFactor();
} else {
factor = new LocalOpentelemetryFactor();
}
return factor?.create();
}
function registerCustomMetrics() {
const host = new HostMetrics({ name: 'instance-host-metrics' });
host.start();
}
export function start() {
const sdk = createSDK();
if (sdk) {
sdk.start();
registerCustomMetrics();
registerBusinessMetrics();
}
}
+21 -99
View File
@@ -1,99 +1,11 @@
import { Counter, Gauge, register, Summary } from 'prom-client';
import { Attributes } from '@opentelemetry/api';
function getOr<T>(name: string, or: () => T): T {
return (register.getSingleMetric(name) as T) || or();
}
type LabelValues<T extends string> = Partial<Record<T, string | number>>;
type MetricsCreator<T extends string> = (
value: number,
labels: LabelValues<T>
) => void;
type TimerMetricsCreator<T extends string> = (
labels: LabelValues<T>
) => () => number;
export const metricsCreatorGenerator = () => {
const counterCreator = <T extends string>(
name: string,
labelNames?: T[]
): MetricsCreator<T> => {
const counter = getOr(
name,
() =>
new Counter({
name,
help: name,
...(labelNames ? { labelNames } : {}),
})
);
return (value: number, labels: LabelValues<T>) => {
counter.inc(labels, value);
};
};
const gaugeCreator = <T extends string>(
name: string,
labelNames?: T[]
): MetricsCreator<T> => {
const gauge = getOr(
name,
() =>
new Gauge({
name,
help: name,
...(labelNames ? { labelNames } : {}),
})
);
return (value: number, labels: LabelValues<T>) => {
gauge.set(labels, value);
};
};
const timerCreator = <T extends string>(
name: string,
labelNames?: T[]
): TimerMetricsCreator<T> => {
const summary = getOr(
name,
() =>
new Summary({
name,
help: name,
...(labelNames ? { labelNames } : {}),
})
);
return (labels: LabelValues<T>) => {
const now = process.hrtime();
return () => {
const delta = process.hrtime(now);
const value = delta[0] + delta[1] / 1e9;
summary.observe(labels, value);
return value;
};
};
};
return {
counter: counterCreator,
gauge: gaugeCreator,
timer: timerCreator,
};
};
export const metricsCreator = metricsCreatorGenerator();
import { getMeter } from './metrics';
export const CallTimer = (
name: string,
labels: Record<string, any> = {}
attrs?: Attributes
): MethodDecorator => {
const timer = metricsCreator.timer(name, Object.keys(labels));
// @ts-expect-error allow
return (
_target,
@@ -106,19 +18,27 @@ export const CallTimer = (
}
desc.value = function (...args: any[]) {
const endTimer = timer(labels);
const timer = getMeter().createHistogram(name, {
description: `function call time costs of ${name}`,
});
const start = Date.now();
const end = () => {
timer.record(Date.now() - start, attrs);
};
let result: any;
try {
result = originalMethod.apply(this, args);
} catch (e) {
endTimer();
end();
throw e;
}
if (result instanceof Promise) {
return result.finally(endTimer);
return result.finally(end);
} else {
endTimer();
end();
return result;
}
};
@@ -129,10 +49,8 @@ export const CallTimer = (
export const CallCounter = (
name: string,
labels: Record<string, any> = {}
attrs?: Attributes
): MethodDecorator => {
const count = metricsCreator.counter(name, Object.keys(labels));
// @ts-expect-error allow
return (
_target,
@@ -145,7 +63,11 @@ export const CallCounter = (
}
desc.value = function (...args: any[]) {
count(1, labels);
const count = getMeter().createCounter(name, {
description: `function call counter of ${name}`,
});
count.add(1, attrs);
return originalMethod.apply(this, args);
};