Node crash with ERR_STREAM_DESTROYED when rotating logs

I was using a simple file transport in my server.

       new winston.transports.File({ filename: C.DEFAULT_LOG_FILENAME })

That has been working fine throughout development. Now that I’m going to production, I added winston-daily-rotate-file, so I replaced the File transport listed above with this:

      new winston.transports.DailyRotateFile({ 
          filename      : C.DEFAULT_LOG_FILENAME + '-%DATE%.log',
          dirname       : './logs',
          datePattern   : 'YYYY-MM-DD:HH',
          zippedArchive : false,
          maxSize       : '2m',
          maxFiles      : '14d',
        })

But I found that my node server was crashing. Turns out it was crashing when it hit the file size limit and rotated log files. I changed the maxSize to “4k” and was able to get it to crash easily by generating log data. Here is the message from the crash:

Error [ERR_STREAM_DESTROYED]: Cannot call write after a stream was destroyed
    at doWrite (_stream_writable.js:406:19)
    at clearBuffer (_stream_writable.js:517:5)
    at onwrite (_stream_writable.js:465:7)
    at fs.write (fs.js:2282:5)
    at FSReqWrap.wrapper [as oncomplete] (fs.js:643:5)

I’m running node version is 10.3.0 on macOS 10.13.5. I’m new to node / winston, so its not clear to me if this is my problem, a problem in node, or a problem in winston.

[shirakaba]

To anyone it may help: I got this error when I ran my app initially with sudo (meaning that logs would be written as root:root), then again without sudo (meaning that logs would be written as ubuntu:ubuntu, and thus lacked privileges to rotate the logs).

So in my case, it wasn’t a fault of Winston. Just a standard consequence of permissions management. To avoid it, always run the app as a user in the same usergroup as you intend to run it with in future.

[mattberther]

I’m moving this issue to #226, which more accurately describes the problem (winston-daily-rotate-file does not work well with process managers).