From 0f59f0e8446eaacd78aad62e2da852940d92965d Mon Sep 17 00:00:00 2001
From: A <258483684+la14-1@users.noreply.github.com>
Date: Sat, 21 Feb 2026 14:00:13 -0800
Subject: [PATCH] fix: fly machine wait timeout exceeds API max of 60s (#1619)

The Fly Machines API enforces a [1s, 1m0s] range on
WaitMachineRequest.Timeout. We were passing 90s, which caused an
invalid_argument error and prevented machines from starting.

Lower the default to 60s (the API maximum) and retry up to 3 times
so slow-starting machines still have a full 3-minute window.

Co-authored-by: lab <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---
 cli/package.json   |  2 +-
 cli/src/fly/fly.ts | 26 ++++++++++++++++++--------
 2 files changed, 19 insertions(+), 9 deletions(-)

diff --git a/cli/package.json b/cli/package.json
index 8d686fb3..b2fe677a 100644
--- a/cli/package.json
+++ b/cli/package.json
@@ -1,6 +1,6 @@
 {
   "name": "@openrouter/spawn",
-  "version": "0.5.21",
+  "version": "0.5.22",
   "type": "module",
   "bin": {
     "spawn": "cli.js"
diff --git a/cli/src/fly/fly.ts b/cli/src/fly/fly.ts
index 4e68be94..add0b39a 100644
--- a/cli/src/fly/fly.ts
+++ b/cli/src/fly/fly.ts
@@ -533,14 +533,25 @@ async function createMachine(
 async function waitForMachineStart(
   name: string,
   machineId: string,
-  timeout = 90,
+  timeout = 60,
+  retries = 3,
 ): Promise<void> {
-  logStep(`Waiting for machine to start (timeout: ${timeout}s)...`);
-  const resp = await flyApi(
-    "GET",
-    `/apps/${name}/machines/${machineId}/wait?state=started&timeout=${timeout}`,
-  );
-  if (hasError(resp)) {
+  for (let attempt = 1; attempt <= retries; attempt++) {
+    logStep(
+      `Waiting for machine to start (timeout: ${timeout}s, attempt ${attempt}/${retries})...`,
+    );
+    const resp = await flyApi(
+      "GET",
+      `/apps/${name}/machines/${machineId}/wait?state=started&timeout=${timeout}`,
+    );
+    if (!hasError(resp)) {
+      logInfo("Machine is running");
+      return;
+    }
+    if (attempt < retries) {
+      logWarn(`Machine not ready yet, retrying...`);
+      continue;
+    }
     const data = parseJson(resp);
     logError(
       `Machine did not reach 'started' state: ${data?.error || "timeout"}`,
@@ -548,7 +559,6 @@ async function waitForMachineStart(
     logError("Try a new region: FLY_REGION=ord spawn fly <agent>");
     throw new Error("Machine start timeout");
   }
-  logInfo("Machine is running");
 }
 
 async function cleanupOnFailure(appName: string): Promise<void> {