fix(docker): healthcheck tries 127.0.0.1/localhost/::1 and surfaces errors (#3151) (#3209)

The healthcheck probed only 127.0.0.1 and swallowed every error (empty
Output), so containers binding to a non-loopback address always reported
unhealthy with no diagnostic. Add a multi-host probe helper that succeeds on
the first 2xx and prints the last error to stderr on total failure.

Co-authored-by: naimo84 <naimo84@users.noreply.github.com>

scripts/dev/healthcheck.mjs

@@ -2,13 +2,83 @@
 
 /**
  * Docker healthcheck script for OmniRoute.
- * Checks the /api/monitoring/health endpoint on the dashboard port.
+ * Probes the /api/monitoring/health endpoint on the dashboard port.
  * Used by Dockerfile and docker-compose files.
+ *
+ * #3151 — in some Docker network setups the server binds to a container IP and
+ * a probe against `127.0.0.1` is not reachable, while `localhost`/`::1` (or vice
+ * versa) is. The previous version probed ONLY `127.0.0.1` and swallowed every
+ * error, so the container was reported `unhealthy` with an empty, undiagnosable
+ * `State.Health[].Output`. We now try an ordered list of hosts and surface the
+ * last error on total failure.
  */
-const port = process.env.DASHBOARD_PORT || process.env.PORT || "20128";
 
-fetch(`http://127.0.0.1:${port}/api/monitoring/health`)
-  .then((r) => {
-    if (!r.ok) throw new Error(`HTTP ${r.status}`);
-  })
-  .catch(() => process.exit(1));
+import { pathToFileURL } from "node:url";
+
+const DEFAULT_HOSTS = ["127.0.0.1", "localhost", "::1"];
+const DEFAULT_TIMEOUT_MS = 4000;
+
+/**
+ * Build the health URL for a host, bracketing IPv6 literals (e.g. `::1`).
+ * @param {string} host
+ * @param {string|number} port
+ */
+function healthUrl(host, port) {
+  const hostPart = host.includes(":") ? `[${host}]` : host;
+  return `http://${hostPart}:${port}/api/monitoring/health`;
+}
+
+/**
+ * Probe the health endpoint across an ordered list of hosts. Resolves with the
+ * first host that returns a 2xx response; rejects with the last error if every
+ * host fails. Each attempt is bounded by a per-host timeout so one unreachable
+ * host cannot hang the whole probe.
+ *
+ * @param {object} opts
+ * @param {string|number} opts.port
+ * @param {string[]} [opts.hosts]
+ * @param {typeof fetch} [opts.fetchImpl]
+ * @param {number} [opts.timeoutMs]
+ * @returns {Promise<string>} the host that succeeded
+ */
+export async function probeHealth({
+  port,
+  hosts = DEFAULT_HOSTS,
+  fetchImpl = fetch,
+  timeoutMs = DEFAULT_TIMEOUT_MS,
+} = {}) {
+  let lastError = new Error("no hosts to probe");
+  for (const host of hosts) {
+    try {
+      const res = await fetchImpl(healthUrl(host, port), {
+        signal: AbortSignal.timeout(timeoutMs),
+      });
+      if (res.ok) return host;
+      lastError = new Error(`${host}: HTTP ${res.status}`);
+    } catch (err) {
+      lastError = new Error(`${host}: ${err instanceof Error ? err.message : String(err)}`);
+    }
+  }
+  throw lastError;
+}
+
+async function main() {
+  const port = process.env.DASHBOARD_PORT || process.env.PORT || "20128";
+  try {
+    await probeHealth({ port });
+    process.exit(0);
+  } catch (err) {
+    // Surface the failure so `docker inspect ... .State.Health[].Output` is
+    // diagnostic instead of empty (#3151).
+    process.stderr.write(`healthcheck failed: ${err instanceof Error ? err.message : err}\n`);
+    process.exit(1);
+  }
+}
+
+// Only auto-run when invoked as the entrypoint (so importing the helper in
+// tests does not trigger a real probe + process.exit).
+const isEntrypoint =
+  Boolean(process.argv[1]) && import.meta.url === pathToFileURL(process.argv[1]).href;
+if (isEntrypoint) {
+  main();
+}

tests/unit/docker-healthcheck-3151.test.ts

@@ -0,0 +1,94 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import http from "node:http";
+
+// #3151 — Docker healthcheck always reports unhealthy because the probe only
+// hit 127.0.0.1 and swallowed every error (empty Output, undiagnosable).
+// The hardened script exports an injectable `probeHealth` helper that tries an
+// ordered host list and surfaces the last error on total failure.
+const { probeHealth } = (await import("../../scripts/dev/healthcheck.mjs")) as {
+  probeHealth: (opts: {
+    port: number | string;
+    hosts?: string[];
+    fetchImpl?: typeof fetch;
+    timeoutMs?: number;
+  }) => Promise<string>;
+};
+
+/** Start an ephemeral HTTP server bound only to the given host. */
+function startServer(host: string): Promise<{ server: http.Server; port: number }> {
+  return new Promise((resolve, reject) => {
+    const server = http.createServer((req, res) => {
+      if (req.url === "/api/monitoring/health") {
+        res.writeHead(200, { "content-type": "application/json" });
+        res.end(JSON.stringify({ status: "ok" }));
+      } else {
+        res.writeHead(404);
+        res.end();
+      }
+    });
+    server.on("error", reject);
+    server.listen(0, host, () => {
+      const addr = server.address();
+      if (addr && typeof addr === "object") {
+        resolve({ server, port: addr.port });
+      } else {
+        reject(new Error("no address"));
+      }
+    });
+  });
+}
+
+function closeServer(server: http.Server): Promise<void> {
+  return new Promise((resolve) => server.close(() => resolve()));
+}
+
+const servers: http.Server[] = [];
+
+test.after(async () => {
+  for (const s of servers) {
+    await closeServer(s);
+  }
+});
+
+test("probeHealth resolves the host when the server is on 127.0.0.1", async () => {
+  const { server, port } = await startServer("127.0.0.1");
+  servers.push(server);
+
+  const ok = await probeHealth({ port, hosts: ["127.0.0.1", "localhost", "::1"] });
+  assert.equal(ok, "127.0.0.1");
+});
+
+test("probeHealth falls through to a later host when 127.0.0.1 is unreachable", async () => {
+  // Bind the real server on 127.0.0.1, but list an unreachable host first so
+  // the helper must fall through. We point the first host at a port with no
+  // listener to simulate ECONNREFUSED, then the real host on the same port.
+  const { server, port } = await startServer("127.0.0.1");
+  servers.push(server);
+
+  // First host resolves to nothing listening (use a host alias that will fail),
+  // second host is the working loopback.
+  const ok = await probeHealth({
+    port,
+    hosts: ["192.0.2.1", "127.0.0.1"], // 192.0.2.1 = TEST-NET-1, unroutable
+    timeoutMs: 300,
+  });
+  assert.equal(ok, "127.0.0.1");
+});
+
+test("probeHealth throws a non-empty error string when every host fails", async () => {
+  // No server started: pick a port unlikely to have a listener.
+  await assert.rejects(
+    () =>
+      probeHealth({
+        port: 1, // privileged/closed port → connection refused
+        hosts: ["127.0.0.1", "localhost"],
+        timeoutMs: 300,
+      }),
+    (err: unknown) => {
+      assert.ok(err instanceof Error);
+      assert.ok(err.message.length > 0, "error message must be non-empty (not swallowed)");
+      return true;
+    }
+  );
+});