cluster模式指定interpreter（自编译的node）不生效。nvm安装的pm2，fork模式单线程是成功的

[bnmgh1]

No description provided.

[rohankamble103]

Problem Explained

In cluster mode, PM2 spawns multiple workers, and it needs the interpreter path (node) to be reliable for every forked child. When you:

Use NVM, the node path is managed inside your user shell session.

Use self-compiled Node, that binary may not be accessible outside the shell or from pm2-runtime (which runs as a different subprocess or in Docker).

So in cluster mode, PM2 fails to find the right node binary — while fork mode just works because it reuses the existing shell/node context.

Solution

1. Explicitly define the Node interpreter path in your ecosystem.config.js

Use the full path to your node binary (e.g., /home/youruser/.nvm/versions/node/vXX.X.X/bin/node or /usr/local/bin/node):

module.exports = {
  apps: [
    {
      name: "app",
      script: "your-entry-file.js",
      exec_mode: "cluster",
      instances: "max",
      interpreter: "/full/path/to/node",
    },
  ]
}

You can get the full path by running:

which node

2. Ensure the environment is available to PM2If you're running pm2 in a Docker container, via pm2-runtime, or as a systemd service, make sure the environment variables and $PATH are properly set.

If using NVM, add this at the top of your Dockerfile or shell profile:

export NVM_DIR="$HOME/.nvm"
[ -s "$NVM_DIR/nvm.sh" ] && \. "$NVM_DIR/nvm.sh"
nvm use 19

Or just point to the Node path directly.

3. Try Node from NVM in Docker/PM2 RuntimeIf you're using NVM inside a Docker container, it's often easier to just install Node using curl or system binaries inside the container to avoid NVM quirks in non-interactive shells.

For example in Docker:

RUN curl -fsSL https://deb.nodesource.com/setup_19.x | bash - \
 && apt-get install -y nodejs

Optional Debug Tip
Add this inside your ecosystem.config.js:

interpreter_args: "--trace-warnings",

To get more logs if cluster mode fails.