fix: make otel tracing leaner and synchronous

package.json

@@ -81,10 +81,10 @@
     "@bull-board/api": "^5.9.1",
     "@bull-board/express": "^5.9.1",
     "@opentelemetry/api": "^1.7.0",
-    "@opentelemetry/auto-instrumentations-node": "^0.44.0",
     "@opentelemetry/exporter-jaeger": "^1.19.0",
     "@opentelemetry/exporter-metrics-otlp-proto": "^0.50.0",
     "@opentelemetry/exporter-trace-otlp-proto": "^0.50.0",
+    "@opentelemetry/instrumentation-http": "^0.53.0",
     "@opentelemetry/resources": "^1.18.1",
     "@opentelemetry/sdk-metrics": "^1.18.1",
     "@opentelemetry/sdk-node": "^0.50.0",

src/controllers/opentelemetry.ts

@@ -1,5 +1,6 @@
 import { NodeSDK } from "@opentelemetry/sdk-node";
-import { getNodeAutoInstrumentations } from "@opentelemetry/auto-instrumentations-node";
+import { SimpleSpanProcessor } from "@opentelemetry/sdk-trace-node";
+import { HttpInstrumentation } from "@opentelemetry/instrumentation-http";
 import { Resource } from "@opentelemetry/resources";
 import Debug from "debug";
 import { OTLPTraceExporter } from '@opentelemetry/exporter-trace-otlp-proto';
@@ -9,17 +10,20 @@ import { ATTR_SERVICE_NAME } from '@opentelemetry/semantic-conventions';
 const jaegerHost = process.env.JAEGER_HOST ?? 'jaeger-otel-collector.sri';
 const jaegerPort = process.env.JAEGER_PORT ?? 4318;
 const jaegerResName = process.env.JAEGER_RES_NAME ?? '/v1/traces';
+const traceExporter = new OTLPTraceExporter({
+  url: `http://${jaegerHost}:${jaegerPort}${jaegerResName}`
+});
 
 debug("Initializing Opentelemetry instrumentation...");
 const sdk = new NodeSDK({
   // metrics, if needed, shall be exported on a different endpoint
-  traceExporter: new OTLPTraceExporter({
-    url: `http://${jaegerHost}:${jaegerPort}${jaegerResName}`
-  }),
-  instrumentations: [getNodeAutoInstrumentations()],
+  // trace a subset of instrumentations to avoid performance overhead
+  instrumentations: [new HttpInstrumentation()],
   resource: new Resource({
     [ATTR_SERVICE_NAME]: "biothings-explorer",
   }),
+  // use simple span processor to avoid losing data when the forked process exits (taskHandler)
+  spanProcessors: [new SimpleSpanProcessor(traceExporter)],
 });
 debug(`OTel URL http://${jaegerHost}:${jaegerPort}${jaegerResName}`);
 sdk.start();

src/controllers/threading/taskHandler.ts

@@ -91,14 +91,13 @@ async function runTask({
     });
 
     let activeContext: Context = propagation.extract(context.active(), { traceparent, tracestate });
-    debug(`OTel task ${traceparent} and ${tracestate}`);
+    debug(`OTel task context: ${traceparent} and ${tracestate}`);
     let tracer = trace.getTracer("biothings-explorer-thread")
     span = tracer.startSpan(
       routeNames[route],
       {kind: 1},  // specifies internal span
       activeContext,
     );
-    debug(`OTel span created ${span}`);
 
     span.setAttribute("bte.requestData", JSON.stringify(req.data.queryGraph));
     Telemetry.setOtelSpan(span);

src/controllers/threading/threadHandler.ts

@@ -1,6 +1,6 @@
 import { MessageChannel, threadId } from "worker_threads";
 import Debug from "debug";
-import { context, propagation } from "@opentelemetry/api";
+import { context, propagation, trace } from "@opentelemetry/api";
 const debug = Debug("bte:biothings-explorer-trapi:threading");
 import path from "path";
 import { redisClient } from "@biothings-explorer/utils";
@@ -221,6 +221,11 @@ export async function runTask(req: Request, res: Response, route: string, useBul
   propagation.inject(context.active(), otelData);
   const { traceparent, tracestate } = otelData;
 
+  // add req dest to root span name as HTTP instrumentation doesn't do it automatically
+  const activeContext = context.active();
+  const rootSpan = trace.getSpan(activeContext);
+  rootSpan.updateName(`${req.method} ${req.originalUrl}`);
+
   const taskInfo: TaskInfo = {
     data: {
       route,