node.js 异步日志记录和未处理的服务器异常的最佳实践

Question

We recently changed our logging code in node.js/Express.js to write to the console asynchronously. However, this comes with an issue where if after the asynchronous code is initiated (and when it runs it will emit a log message), the synchronous code run after that initiation crashes the server, then the logs are never written.

Is there an established pattern or best practice for handling this so that disk I/O is not a blocker for logging, but in the event of an unhandled exception (such as an exception occurring in our last-chance exception handler in Express.js), the logs will still be written to disk/console?

I thought about starting a separate logging process and doing inter-process communication, but that feels pretty darn heavy just to avoid a little bit of I/O. And I don't think a separate thread will work—won't the node.js unhandled exception that takes the server down also kill all threads started/owned by the process? (I'm no expert in this area.)

Or is it better to just write to the console synchronously in our logger, and thereby prevent this situation in the first place?

Answer 1

Your changed logging code should give you access to the underlying stream. The synchronous code that crashes the server likely does it by throwing an uncaught exception. Which you can catch using

process.on('uncaughtException', err => { logSync(err); exit(errCode); })

Then log the exception synchronously using the stream or console and exit.