refactor(server): telemetry (#13588)

refactor: telemetry
This commit is contained in:
Jason Rasmussen
2024-10-21 19:52:30 -04:00
committed by GitHub
parent 16f2364e93
commit e6a666f1d3
45 changed files with 143 additions and 202 deletions

View File

@@ -1,7 +1,22 @@
import { Inject, Injectable } from '@nestjs/common';
import { Reflector } from '@nestjs/core';
import { MetricOptions } from '@opentelemetry/api';
import { AsyncLocalStorageContextManager } from '@opentelemetry/context-async-hooks';
import { PrometheusExporter } from '@opentelemetry/exporter-prometheus';
import { HttpInstrumentation } from '@opentelemetry/instrumentation-http';
import { IORedisInstrumentation } from '@opentelemetry/instrumentation-ioredis';
import { NestInstrumentation } from '@opentelemetry/instrumentation-nestjs-core';
import { PgInstrumentation } from '@opentelemetry/instrumentation-pg';
import { NodeSDK, contextBase, metrics, resources } from '@opentelemetry/sdk-node';
import { SemanticResourceAttributes } from '@opentelemetry/semantic-conventions';
import { ClassConstructor } from 'class-transformer';
import { snakeCase, startCase } from 'lodash';
import { MetricService } from 'nestjs-otel';
import { copyMetadataFromFunctionToFunction } from 'nestjs-otel/lib/opentelemetry.utils';
import { serverVersion } from 'src/constants';
import { MetadataKey } from 'src/enum';
import { IConfigRepository } from 'src/interfaces/config.interface';
import { ILoggerRepository } from 'src/interfaces/logger.interface';
import { IMetricGroupRepository, ITelemetryRepository, MetricGroupOptions } from 'src/interfaces/telemetry.interface';
class MetricGroupRepository implements IMetricGroupRepository {
@@ -33,6 +48,43 @@ class MetricGroupRepository implements IMetricGroupRepository {
}
}
const aggregation = new metrics.ExplicitBucketHistogramAggregation(
[0.1, 0.25, 0.5, 0.75, 1, 2.5, 5, 7.5, 10, 25, 50, 75, 100, 250, 500, 750, 1000, 2500, 5000, 7500, 10_000],
true,
);
let instance: NodeSDK | undefined;
export const bootstrapTelemetry = (port: number) => {
if (instance) {
throw new Error('OpenTelemetry SDK already started');
}
instance = new NodeSDK({
resource: new resources.Resource({
[SemanticResourceAttributes.SERVICE_NAME]: `immich`,
[SemanticResourceAttributes.SERVICE_VERSION]: serverVersion.toString(),
}),
metricReader: new PrometheusExporter({ port }),
contextManager: new AsyncLocalStorageContextManager(),
instrumentations: [
new HttpInstrumentation(),
new IORedisInstrumentation(),
new NestInstrumentation(),
new PgInstrumentation(),
],
views: [new metrics.View({ aggregation, instrumentName: '*', instrumentUnit: 'ms' })],
});
instance.start();
};
export const teardownTelemetry = async () => {
if (instance) {
await instance.shutdown();
instance = undefined;
}
};
@Injectable()
export class TelemetryRepository implements ITelemetryRepository {
api: MetricGroupRepository;
@@ -40,8 +92,13 @@ export class TelemetryRepository implements ITelemetryRepository {
jobs: MetricGroupRepository;
repo: MetricGroupRepository;
constructor(metricService: MetricService, @Inject(IConfigRepository) configRepository: IConfigRepository) {
const { telemetry } = configRepository.getEnv();
constructor(
private metricService: MetricService,
private reflect: Reflector,
@Inject(IConfigRepository) private configRepository: IConfigRepository,
@Inject(ILoggerRepository) private logger: ILoggerRepository,
) {
const { telemetry } = this.configRepository.getEnv();
const { apiMetrics, hostMetrics, jobMetrics, repoMetrics } = telemetry;
this.api = new MetricGroupRepository(metricService).configure({ enabled: apiMetrics });
@@ -49,4 +106,61 @@ export class TelemetryRepository implements ITelemetryRepository {
this.jobs = new MetricGroupRepository(metricService).configure({ enabled: jobMetrics });
this.repo = new MetricGroupRepository(metricService).configure({ enabled: repoMetrics });
}
setup({ repositories }: { repositories: ClassConstructor<unknown>[] }) {
const { telemetry } = this.configRepository.getEnv();
if (!telemetry.enabled || !telemetry.repoMetrics) {
return;
}
for (const Repository of repositories) {
const isEnabled = this.reflect.get(MetadataKey.TELEMETRY_ENABLED, Repository) ?? true;
if (!isEnabled) {
this.logger.debug(`Telemetry disabled for ${Repository.name}`);
continue;
}
this.wrap(Repository);
}
}
private wrap(Repository: ClassConstructor<unknown>) {
const className = Repository.name;
const descriptors = Object.getOwnPropertyDescriptors(Repository.prototype);
const unit = 'ms';
for (const [propName, descriptor] of Object.entries(descriptors)) {
const isMethod = typeof descriptor.value == 'function' && propName !== 'constructor';
if (!isMethod) {
continue;
}
const method = descriptor.value;
const propertyName = snakeCase(String(propName));
const metricName = `${snakeCase(className).replaceAll(/_(?=(repository)|(controller)|(provider)|(service)|(module))/g, '.')}.${propertyName}.duration`;
const histogram = this.metricService.getHistogram(metricName, {
prefix: 'immich',
description: `The elapsed time in ${unit} for the ${startCase(className)} to ${propertyName.toLowerCase()}`,
unit,
valueType: contextBase.ValueType.DOUBLE,
});
descriptor.value = function (...args: any[]) {
const start = performance.now();
const result = method.apply(this, args);
void Promise.resolve(result)
.then(() => histogram.record(performance.now() - start, {}))
.catch(() => {
// noop
});
return result;
};
copyMetadataFromFunctionToFunction(method, descriptor.value);
Object.defineProperty(Repository.prototype, propName, descriptor);
}
}
}