@roamcode.ai/server 1.0.23 → 1.1.0

package/dist/health-watchdog.js

@@ -0,0 +1,87 @@
+// src/health-watchdog.ts
+import { spawn } from "child_process";
+import { fileURLToPath } from "url";
+var HEALTH_INSTANCE_HEADER = "x-roamcode-instance";
+var WATCHDOG_INITIAL_DELAY_MS = 5e3;
+var WATCHDOG_INTERVAL_MS = 1e4;
+var WATCHDOG_PROBE_TIMEOUT_MS = 3e3;
+var WATCHDOG_FAILURE_THRESHOLD = 4;
+var WATCHDOG_TERMINATION_GRACE_MS = 1e4;
+function boundedInteger(value, min, max) {
+  if (!value || !/^\d+$/.test(value)) return void 0;
+  const parsed = Number(value);
+  return Number.isSafeInteger(parsed) && parsed >= min && parsed <= max ? parsed : void 0;
+}
+function parseHealthWatchdogEnv(env) {
+  const parentPid = boundedInteger(env.ROAMCODE_WATCHDOG_PARENT_PID, 2, 2 ** 31 - 1);
+  const port = boundedInteger(env.ROAMCODE_WATCHDOG_PORT, 1, 65535);
+  const instanceId = env.ROAMCODE_WATCHDOG_INSTANCE_ID;
+  if (!parentPid || !port || !instanceId || !/^[A-Za-z0-9_-]{16,128}$/.test(instanceId)) return void 0;
+  return { parentPid, port, instanceId };
+}
+async function defaultProbe(config2) {
+  try {
+    const response = await fetch(`http://127.0.0.1:${config2.port}/health`, {
+      cache: "no-store",
+      headers: { connection: "close" },
+      signal: AbortSignal.timeout(config2.probeTimeoutMs ?? WATCHDOG_PROBE_TIMEOUT_MS)
+    });
+    const matches = response.status === 200 && response.headers.get(HEALTH_INSTANCE_HEADER) === config2.instanceId;
+    await response.body?.cancel().catch(() => void 0);
+    return matches;
+  } catch {
+    return false;
+  }
+}
+function defaultParentAlive(parentPid) {
+  if (process.ppid !== parentPid) return false;
+  try {
+    process.kill(parentPid, 0);
+    return true;
+  } catch (error) {
+    return error.code === "EPERM";
+  }
+}
+var defaultSleep = (milliseconds) => new Promise((resolve) => setTimeout(resolve, milliseconds));
+async function runHealthWatchdog(config2, deps = {}) {
+  const probe = deps.probe ?? (() => defaultProbe(config2));
+  const parentAlive = deps.parentAlive ?? (() => defaultParentAlive(config2.parentPid));
+  const terminateParent = deps.terminateParent ?? ((signal) => {
+    try {
+      process.kill(config2.parentPid, signal);
+    } catch {
+    }
+  });
+  const sleep = deps.sleep ?? defaultSleep;
+  const now = deps.now ?? Date.now;
+  const log = deps.log ?? ((message) => process.stderr.write(`${message}
+`));
+  const threshold = Math.max(1, config2.failureThreshold ?? WATCHDOG_FAILURE_THRESHOLD);
+  const intervalMs = config2.intervalMs ?? WATCHDOG_INTERVAL_MS;
+  const graceMs = config2.terminationGraceMs ?? WATCHDOG_TERMINATION_GRACE_MS;
+  let consecutiveFailures = 0;
+  await sleep(config2.initialDelayMs ?? WATCHDOG_INITIAL_DELAY_MS);
+  while (parentAlive()) {
+    const healthy = await probe().catch(() => false);
+    consecutiveFailures = healthy ? 0 : consecutiveFailures + 1;
+    if (consecutiveFailures >= threshold) {
+      log(
+        `[roamcode] health watchdog observed ${consecutiveFailures} consecutive failed probes; restarting the unresponsive server`
+      );
+      terminateParent("SIGTERM");
+      const deadline = now() + graceMs;
+      while (parentAlive() && now() < deadline) await sleep(Math.min(250, Math.max(1, deadline - now())));
+      if (parentAlive()) terminateParent("SIGKILL");
+      return;
+    }
+    await sleep(intervalMs);
+  }
+}
+
+// src/health-watchdog-entry.ts
+var config = parseHealthWatchdogEnv(process.env);
+if (!config) process.exit(2);
+void runHealthWatchdog(config).then(
+  () => process.exit(0),
+  () => process.exit(1)
+);