fix(server): refine error handling and PM2 connection logic

- Reduced the maximum restarts for PM2 processes from 20 to 8 and added a restart delay of 2000ms for better stability.
- Improved error handling during server startup and PM2 connection, ensuring clearer logging and graceful exits on failure.
- Introduced a utility function to filter out PM2 metadata from environment variables, preventing conflicts during process management.

ecosystem.config.cjs

@@ -48,7 +48,8 @@ module.exports = {
       instances: 1,
       exec_mode: "fork",
       autorestart: true,
-      max_restarts: 20,
+      max_restarts: 8,
+      restart_delay: 2000,
       env: {
         ...env,
         JFLOW_CONTROL_PORT: env.JFLOW_CONTROL_PORT ?? "8600",

packages/server/control.js

@@ -64,13 +64,6 @@ try {
   process.exit(1);
 }
 
-try {
-  await connectPm2();
-} catch (err) {
-  log.error({ err }, "failed to connect to PM2 — is pm2 installed?");
-  process.exit(1);
-}
-
 async function applyDesiredState() {
   const state = readControlState();
   await ensureHttp({
@@ -98,8 +91,6 @@ async function applyDesiredState() {
   );
 }
 
-await applyDesiredState();
-
 const server = fastify({ loggerInstance: log });
 await server.register(cookie);
 await server.register(jwt, {
@@ -484,12 +475,24 @@ process.on("SIGINT", shutdown);
 process.on("SIGTERM", shutdown);
 
 const port = Number(process.env.JFLOW_CONTROL_PORT ?? process.env.PORT ?? 8600);
-server
+try {
-  .listen({ host: "0.0.0.0", port })
+  await server.listen({ host: "0.0.0.0", port });
-  .then(() => {
   log.info(`Control is running on port ${port}`);
-  })
+} catch (err) {
-  .catch((err) => {
   log.error({ err }, "failed to start control");
   process.exit(1);
-  });
+}
+
+try {
+  await connectPm2();
+} catch (err) {
+  log.error({ err }, "failed to connect to PM2 — is pm2 installed?");
+  process.exit(1);
+}
+
+try {
+  await applyDesiredState();
+} catch (err) {
+  log.error({ err }, "failed to apply desired state");
+  process.exit(1);
+}

packages/server/pm2-bridge.js

@@ -164,13 +164,57 @@ export async function restartPm2Process(pmId) {
   return { name: proc.name, pmId: id };
 }
 
+/**
+ * PM2 injects these into process.env of a managed app. Spreading them into
+ * `pm2.start({ env })` overwrites `name` / `pm_exec_path` so God restarts
+ * jflow-control instead of launching http/worker (EADDRINUSE :8600 loop).
+ */
+const PM2_META_KEYS = new Set([
+  "name",
+  "namespace",
+  "exec_mode",
+  "exec_interpreter",
+  "instances",
+  "instance_var",
+  "node_app_instance",
+  "unique_id",
+  "status",
+  "username",
+  "windowsHide",
+  "merge_logs",
+  "vizion",
+  "vizion_running",
+  "autostart",
+  "autorestart",
+  "automation",
+  "km_link",
+]);
+
+/**
+ * @param {NodeJS.ProcessEnv} env
+ * @returns {NodeJS.ProcessEnv}
+ */
+export function withoutPm2Meta(env) {
+  /** @type {NodeJS.ProcessEnv} */
+  const out = {};
+  for (const [key, val] of Object.entries(env)) {
+    if (val == null) continue;
+    if (PM2_META_KEYS.has(key)) continue;
+    if (key.startsWith("pm_") || key.startsWith("axm_") || key.startsWith("PM2_")) {
+      continue;
+    }
+    out[key] = val;
+  }
+  return out;
+}
+
 /**
  * Shared env for child processes.
  * @param {{ generation: number }} opts
  */
 export function childEnv(opts) {
   return {
-    ...process.env,
+    ...withoutPm2Meta(process.env),
     JFLOW_CONFIG_GENERATION: String(opts.generation),
     JFLOW_CORS_ORIGIN: process.env.JFLOW_CORS_ORIGIN ?? "http://localhost:8500",
     PORT: process.env.JFLOW_HTTP_PORT ?? "8700",