From 01b81fe1441d83ea4bbb913f1e76eded79d19384 Mon Sep 17 00:00:00 2001
From: Intern Dev <dev@wukongai.io>
Date: Mon, 13 Apr 2026 00:48:59 -0400
Subject: [PATCH] fix(nim): bound curl health probes with connect and max
 timeouts

The NIM health probe loop and nimStatusByName() use `curl -sf` without
any timeout. On a hung or silently-dropping NIM endpoint the curl call
can block indefinitely, stalling onboard/waitForNimHealth and CLI status.

Add `--connect-timeout 5 --max-time 5` to both probe calls so each
attempt fails fast on unresponsive endpoints and the outer retry loop
can make progress.

Signed-off-by: Intern Dev <dev@wukongai.io>
---
 src/lib/nim.test.ts | 44 +++++++++++++++++++++++++++++++++++++++-----
 src/lib/nim.ts      | 25 +++++++++++++++++++++----
 2 files changed, 60 insertions(+), 9 deletions(-)

diff --git a/src/lib/nim.test.ts b/src/lib/nim.test.ts
index 85edca72308..a8db9e19477 100644
--- a/src/lib/nim.test.ts
+++ b/src/lib/nim.test.ts
@@ -32,6 +32,24 @@ function loadNimWithMockedRunner(runCapture: Mock) {
   };
 }
 
+/** Check if an argv array or legacy shell command contains a specific argument. */
+function hasArg(cmd: string | string[], arg: string): boolean {
+  return Array.isArray(cmd) ? cmd.includes(arg) : cmd.includes(arg);
+}
+
+function hasCurlTimeoutArgs(cmd: string | string[]): boolean {
+  if (!Array.isArray(cmd)) {
+    return (
+      cmd.includes("curl") &&
+      cmd.includes("--connect-timeout 5") &&
+      cmd.includes("--max-time 5")
+    );
+  }
+  const connectTimeout = cmd.indexOf("--connect-timeout");
+  const maxTime = cmd.indexOf("--max-time");
+  return cmd[0] === "curl" && cmd[connectTimeout + 1] === "5" && cmd[maxTime + 1] === "5";
+}
+
 describe("nim", () => {
   describe("listModels", () => {
     it("returns 5 models", () => {
@@ -176,12 +194,27 @@ describe("nim", () => {
     });
   });
 
-  describe("nimStatusByName", () => {
-    /** Check if an argv array contains a specific element. */
-    function hasArg(cmd: string | string[], arg: string): boolean {
-      return Array.isArray(cmd) ? cmd.includes(arg) : cmd.includes(arg);
-    }
+  describe("waitForNimHealth", () => {
+    it("bounds curl health probes with connect and total timeouts", () => {
+      const runCapture = vi.fn((cmd: string | string[]) => {
+        if (!Array.isArray(cmd)) throw new Error("expected argv array");
+        if (cmd[0] === "curl" && hasArg(cmd, "http://localhost:9000/v1/models")) return '{"data":[]}';
+        return "";
+      });
+      const { nimModule, restore } = loadNimWithMockedRunner(runCapture);
 
+      try {
+        expect(nimModule.waitForNimHealth(9000, 1)).toBe(true);
+        const commands = runCapture.mock.calls.map(([c]: [string | string[]]) => c);
+
+        expect(commands.some((c) => c[0] === "curl" && hasCurlTimeoutArgs(c))).toBe(true);
+      } finally {
+        restore();
+      }
+    });
+  });
+
+  describe("nimStatusByName", () => {
     it("uses provided port directly", () => {
       const runCapture = vi.fn((cmd: string | string[]) => {
         if (!Array.isArray(cmd)) throw new Error("expected argv array");
@@ -205,6 +238,7 @@ describe("nim", () => {
         expect(commands.some((c) => c.includes("http://localhost:9000/v1/models"))).toBe(
           true,
         );
+        expect(commands.some((c) => c[0] === "curl" && hasCurlTimeoutArgs(c))).toBe(true);
       } finally {
         restore();
       }
diff --git a/src/lib/nim.ts b/src/lib/nim.ts
index 03f17f4ba24..2a3da454d77 100644
--- a/src/lib/nim.ts
+++ b/src/lib/nim.ts
@@ -220,9 +220,18 @@ export function waitForNimHealth(port = VLLM_PORT, timeout = 300): boolean {
 
   while ((Date.now() - start) / 1000 < timeout) {
     try {
-      const result = runCapture(["curl", "-sf", `http://localhost:${hostPort}/v1/models`], {
-        ignoreError: true,
-      });
+      const result = runCapture(
+        [
+          "curl",
+          "-sf",
+          "--connect-timeout",
+          "5",
+          "--max-time",
+          "5",
+          `http://localhost:${hostPort}/v1/models`,
+        ],
+        { ignoreError: true },
+      );
       if (result) {
         console.log("  NIM is healthy.");
         return true;
@@ -272,7 +281,15 @@ export function nimStatusByName(name: string, port?: number): NimStatus {
         resolvedHostPort = m ? Number(m[1]) : VLLM_PORT;
       }
       const health = runCapture(
-        ["curl", "-sf", `http://localhost:${resolvedHostPort}/v1/models`],
+        [
+          "curl",
+          "-sf",
+          "--connect-timeout",
+          "5",
+          "--max-time",
+          "5",
+          `http://localhost:${resolvedHostPort}/v1/models`,
+        ],
         { ignoreError: true },
       );
       healthy = !!health;