Refactor OpenTelemetry integration and cleanup

- Removed deprecated OpenTelemetry SDK dependencies from package.json and pnpm-lock.yaml.
- Updated instrumentation registration to utilize the new @vercel/otel package for improved performance and compatibility.
- Enhanced error handling in backend URL configuration to ensure valid JSON parsing.
- Cleaned up unused shutdownOTel function and related code in the Prisma client.

These changes streamline the OpenTelemetry setup and improve the overall resilience of the backend API.

Author: mantrakp04

.github/workflows/e2e-fallback-tests.yaml

@@ -91,7 +91,7 @@ jobs:
       - name: Start stack-backend on fallback port (8110)
         uses: JarvusInnovations/background-action@v1.0.7
         with:
-          run: pnpm -C apps/backend run with-env:test next start --port 8110 --log-order=stream &
+          run: pnpm -C apps/backend run with-env:test next start --port 8110 &
           wait-on: |
             http://localhost:8110
           tail: true

apps/backend/package.json

@@ -69,7 +69,6 @@
     "@opentelemetry/instrumentation": "^0.53.0",
     "@opentelemetry/resources": "^1.26.0",
     "@opentelemetry/sdk-logs": "^0.53.0",
-    "@opentelemetry/sdk-node": "^0.214.0",
     "@opentelemetry/sdk-trace-base": "^1.26.0",
     "@opentelemetry/sdk-trace-node": "^1.26.0",
     "@opentelemetry/semantic-conventions": "^1.27.0",

apps/backend/src/app/api/latest/internal/backend-urls/route.tsx

@@ -47,9 +47,17 @@ let cachedEntries: ReturnType<typeof parseAndValidateConfig> | undefined;
 function getCachedConfig() {
   if (!cachedEntries) {
     const rawEnv = getEnvVariable("STACK_BACKEND_URLS_CONFIG", "");
-    cachedEntries = rawEnv
-      ? parseAndValidateConfig(JSON.parse(rawEnv))
-      : [{ probability: 1, urls: getDefaultApiUrls(getEnvVariable("NEXT_PUBLIC_STACK_API_URL")) }];
+    if (rawEnv) {
+      let parsed;
+      try {
+        parsed = JSON.parse(rawEnv);
+      } catch (e) {
+        throw new StackAssertionError(`STACK_BACKEND_URLS_CONFIG is not valid JSON: ${e}`);
+      }
+      cachedEntries = parseAndValidateConfig(parsed);
+    } else {
+      cachedEntries = [{ probability: 1, urls: getDefaultApiUrls(getEnvVariable("NEXT_PUBLIC_STACK_API_URL")) }];
+    }
   }
   return cachedEntries;
 }

apps/backend/src/instrumentation.ts

@@ -5,70 +5,31 @@ import * as Sentry from "@sentry/nextjs";
 import { getEnvVariable, getNextRuntime, getNodeEnvironment } from "@stackframe/stack-shared/dist/utils/env";
 import { sentryBaseConfig } from "@stackframe/stack-shared/dist/utils/sentry";
 import { nicify } from "@stackframe/stack-shared/dist/utils/strings";
+import { registerOTel } from '@vercel/otel';
 import { initPerfStats } from "./lib/dev-perf-stats";
 import "./polyfills";
 
 // this is a hack for making prisma instrumentation work
 // somehow prisma instrumentation accesses global and it makes edge instrumentation complain
 globalThis.global = globalThis;
 
-function getOTelInstrumentations() {
-  return [
-    new PrismaInstrumentation(),
-    ...getNextRuntime() === "nodejs" ? getNodeAutoInstrumentations({
-      '@opentelemetry/instrumentation-http': {
-        enabled: false,
-      },
-    }) : [],
-  ];
-}
-
-function getDevTraceExporter() {
-  if (getNodeEnvironment() === "development" && getNextRuntime() === "nodejs") {
-    return new OTLPTraceExporter({
-      url: `http://localhost:${getEnvVariable("NEXT_PUBLIC_STACK_PORT_PREFIX", "81")}31/v1/traces`,
-    });
-  }
-  return undefined;
-}
-
-let otelSdk: { shutdown(): Promise<void> } | undefined;
-
-export async function shutdownOTel() {
-  await otelSdk?.shutdown();
-}
-
-async function registerOTelProvider() {
-  const instrumentations = getOTelInstrumentations();
-  const devExporter = getDevTraceExporter();
-
-  if (getEnvVariable("VERCEL", "")) {
-    // On Vercel: use @vercel/otel which wraps the standard OTEL SDK with Vercel-specific defaults
-    const { registerOTel } = await import("@vercel/otel");
-    registerOTel({
-      serviceName: 'stack-backend',
-      instrumentations,
-      ...devExporter ? { traceExporter: devExporter } : {},
-    });
-  } else if (getNextRuntime() === "nodejs") {
-    // On Cloud Run / self-hosted: use standard @opentelemetry/sdk-node (Node.js only)
-    const { NodeSDK } = await import("@opentelemetry/sdk-node");
-    const otelEndpoint = getEnvVariable("OTEL_EXPORTER_OTLP_ENDPOINT", "");
-    const exporter = devExporter ?? (otelEndpoint ? new OTLPTraceExporter({ url: otelEndpoint }) : undefined);
-    const sdk = new NodeSDK({
-      serviceName: 'stack-backend',
-      instrumentations,
-      // Cast needed: @opentelemetry/exporter-trace-otlp-http may be a different major than sdk-node,
-      // but the runtime interface is compatible
-      ...(exporter ? { traceExporter: exporter as any } : {}),
-    });
-    sdk.start();
-    otelSdk = sdk;
-  }
-}
-
 export async function register() {
-  await registerOTelProvider();
+  registerOTel({
+    serviceName: 'stack-backend',
+    instrumentations: [
+      new PrismaInstrumentation(),
+      ...getNextRuntime() === "nodejs" ? getNodeAutoInstrumentations({
+        '@opentelemetry/instrumentation-http': {
+          enabled: false,
+        },
+      }) : [],
+    ],
+    ...getNodeEnvironment() === "development" && getNextRuntime() === "nodejs" ? {
+      traceExporter: new OTLPTraceExporter({
+        url: `http://localhost:${getEnvVariable("NEXT_PUBLIC_STACK_PORT_PREFIX", "81")}31/v1/traces`,
+      }),
+    } : {},
+  });
 
   if (getNextRuntime() === "nodejs") {
     (globalThis as any).process.title = `stack-backend:${getEnvVariable("NEXT_PUBLIC_STACK_PORT_PREFIX", "81")} (node/nextjs)`;

apps/backend/src/prisma-client.tsx

@@ -18,7 +18,6 @@ import net from "node:net";
 import { Pool } from "pg";
 import { isPromise } from "util/types";
 import { runMigrationNeeded } from "./auto-migrations";
-import { shutdownOTel } from "./instrumentation";
 import { registerPgPool } from "./lib/dev-perf-stats";
 import { Tenancy } from "./lib/tenancies";
 import { ensurePolyfilled } from "./polyfills";
@@ -116,7 +115,6 @@ if (!getEnvVariable("VERCEL", "") && !globalVar.__stack_prisma_sigterm_registere
       try {
         console.log("[SIGTERM] Draining background tasks and database connections...");
         await drainInFlightPromises(8000);
-        await shutdownOTel();
         for (const [, entry] of postgresPrismaClientsStore) {
           await entry.client.$disconnect();
         }
@@ -126,6 +124,7 @@ if (!getEnvVariable("VERCEL", "") && !globalVar.__stack_prisma_sigterm_registere
         console.log("[SIGTERM] Completed draining background tasks and database connections.");
       } finally {
         clearTimeout(keepAlive);
+        process.exit(0);
       }
     });
   });

packages/stack-shared/src/interface/client-interface.ts

@@ -181,7 +181,7 @@ export class StackClientInterface {
         } catch (probeError) {
           if (probeError instanceof KnownError) throw probeError;
           // Still down — reduce probe frequency
-          this._currentProbeRate *= 0.5;
+          this._currentProbeRate = Math.max(this._currentProbeRate * 0.5, 0.01);
         }
       }
       return await cb(apiUrls[activeIndex], { maxAttempts: 1, skipDiagnostics: false });