Merge pull request #893 from pikasTech/fix/v02-metrics-sidecar-http-probe

fix: use HTTP request for v02 metrics health probes

internal/dev-entrypoint/metrics-sidecar.mjs

@@ -1,4 +1,5 @@
-import { createServer } from "node:http";
+import { createServer, request as httpRequest } from "node:http";
+import { request as httpsRequest } from "node:https";
 import { realpathSync } from "node:fs";
 import { pathToFileURL } from "node:url";
 
@@ -23,26 +24,49 @@ function metricLabels(env) {
   return `service="${escapeLabelValue(service)}",namespace="${escapeLabelValue(namespace)}",gitops_target="${escapeLabelValue(gitopsTarget)}"`;
 }
 
-export async function probeTarget({ targetUrl, timeoutMs = DEFAULT_TARGET_TIMEOUT_MS, fetchImpl = globalThis.fetch } = {}) {
+export function requestTarget({ targetUrl, timeoutMs = DEFAULT_TARGET_TIMEOUT_MS } = {}) {
   const configured = typeof targetUrl === "string" && targetUrl.trim().length > 0;
-  if (!configured || typeof fetchImpl !== "function") {
+  if (!configured) return Promise.resolve({ configured: false, success: false, statusCode: 0, durationSeconds: 0 });
+
+  let url;
+  try {
+    url = new URL(targetUrl);
+  } catch {
+    return Promise.resolve({ configured: true, success: false, statusCode: 0, durationSeconds: 0 });
+  }
+
+  const requestImpl = url.protocol === "https:" ? httpsRequest : url.protocol === "http:" ? httpRequest : null;
+  if (!requestImpl) return Promise.resolve({ configured: true, success: false, statusCode: 0, durationSeconds: 0 });
+
+  const started = Date.now();
+  return new Promise((resolve) => {
+    const finish = (result) => resolve({
+      configured: true,
+      success: false,
+      statusCode: 0,
+      durationSeconds: (Date.now() - started) / 1000,
+      ...result
+    });
+    const request = requestImpl(url, { method: "GET", timeout: timeoutMs }, (response) => {
+      response.resume();
+      response.on("end", () => finish({ success: Boolean(response.statusCode && response.statusCode >= 200 && response.statusCode < 300), statusCode: response.statusCode || 0 }));
+    });
+    request.on("timeout", () => request.destroy());
+    request.on("error", () => finish({}));
+    request.end();
+  });
+}
+
+export async function probeTarget({ targetUrl, timeoutMs = DEFAULT_TARGET_TIMEOUT_MS, probeImpl = requestTarget } = {}) {
+  const configured = typeof targetUrl === "string" && targetUrl.trim().length > 0;
+  if (!configured || typeof probeImpl !== "function") {
     return { configured: false, success: false, statusCode: 0, durationSeconds: 0 };
   }
-  const controller = new AbortController();
-  const started = Date.now();
-  const timeout = setTimeout(() => controller.abort(), timeoutMs);
   try {
-    const response = await fetchImpl(targetUrl, { signal: controller.signal });
-    return {
-      configured: true,
-      success: response.ok,
-      statusCode: response.status || 0,
-      durationSeconds: (Date.now() - started) / 1000
-    };
+    return await probeImpl({ targetUrl, timeoutMs });
   } catch {
+    const started = Date.now();
     return { configured: true, success: false, statusCode: 0, durationSeconds: (Date.now() - started) / 1000 };
-  } finally {
-    clearTimeout(timeout);
   }
 }
 
@@ -85,7 +109,7 @@ function sendJson(response, statusCode, body) {
   response.end(payload);
 }
 
-export function createMetricsSidecarServer({ env = process.env, fetchImpl = globalThis.fetch } = {}) {
+export function createMetricsSidecarServer({ env = process.env, probeImpl = requestTarget } = {}) {
   return createServer(async (request, response) => {
     const url = new URL(request.url || "/", "http://hwlab-metrics.local");
     if (request.method === "GET" && (url.pathname === "/health" || url.pathname === "/health/live")) {
@@ -98,7 +122,7 @@ export function createMetricsSidecarServer({ env = process.env, fetchImpl = glob
     }
     if (request.method === "GET" && url.pathname === "/metrics") {
       const timeoutMs = Number.parseInt(env.HWLAB_METRICS_TARGET_TIMEOUT_MS || "", 10) || DEFAULT_TARGET_TIMEOUT_MS;
-      const probe = await probeTarget({ targetUrl: env.HWLAB_METRICS_TARGET_URL, timeoutMs, fetchImpl });
+      const probe = await probeTarget({ targetUrl: env.HWLAB_METRICS_TARGET_URL, timeoutMs, probeImpl });
       const payload = buildMetricsText({ env, probe });
       response.writeHead(200, {
         "content-type": "text/plain; version=0.0.4; charset=utf-8",

internal/dev-entrypoint/metrics-sidecar.test.mjs

@@ -1,9 +1,11 @@
 import assert from "node:assert/strict";
 import test from "node:test";
 
+import { createServer } from "node:http";
+import { once } from "node:events";
 import { pathToFileURL } from "node:url";
 
-import { buildMetricsText, isMainModule, sanitizeLabelValue } from "./metrics-sidecar.mjs";
+import { buildMetricsText, isMainModule, probeTarget, requestTarget, sanitizeLabelValue } from "./metrics-sidecar.mjs";
 
 test("metrics text exposes stable HWLAB labels without sensitive identifiers", () => {
   const text = buildMetricsText({
@@ -26,6 +28,40 @@ test("label sanitization keeps metrics labels low-risk", () => {
   assert.equal(sanitizeLabelValue(""), "unknown");
 });
 
+test("health target probe uses Node HTTP request against cluster-local style targets", async () => {
+  const server = createServer((request, response) => {
+    if (request.url === "/health/live") {
+      response.writeHead(200, { "content-type": "application/json" });
+      response.end(JSON.stringify({ status: "ok" }));
+      return;
+    }
+    response.writeHead(404);
+    response.end();
+  });
+  server.listen(0, "127.0.0.1");
+  await once(server, "listening");
+  try {
+    const address = server.address();
+    const probe = await requestTarget({ targetUrl: `http://127.0.0.1:${address.port}/health/live`, timeoutMs: 1000 });
+    assert.equal(probe.configured, true);
+    assert.equal(probe.success, true);
+    assert.equal(probe.statusCode, 200);
+    assert.ok(probe.durationSeconds >= 0);
+  } finally {
+    server.close();
+    await once(server, "close");
+  }
+});
+
+test("health target probe degrades invalid URLs to failed metrics", async () => {
+  const probe = await probeTarget({ targetUrl: "not a url", timeoutMs: 1000 });
+  assert.deepEqual({ configured: probe.configured, success: probe.success, statusCode: probe.statusCode }, {
+    configured: true,
+    success: false,
+    statusCode: 0
+  });
+});
+
 test("entrypoint detection follows ConfigMap symlink paths", () => {
   const realScriptPath = "/var/run/configmaps/hwlab-v02-metrics-sidecar/..2026_06_05/metrics-sidecar.mjs";
   const mountedScriptPath = "/metrics/metrics-sidecar.mjs";