Docker exit status 1 for Node app on AWS - node.js

Question

I'm hosting a beta app on AWS using Express.js, Node, mongoose and docker. Daily active users < 10, mainly friends of mine for testing. The app is down almost everyday for some reason. Initially I thought it was AWS's issue, so I stopped my app, changed it from free tier to t2.medium and started it again.
It didn't resolve the issue, I checked docker log for the container. It was not caused by OOMKilled.
"State": {
"Status": "exited",
"Running": false,
"Paused": false,
"Restarting": false,
"OOMKilled": false,
"Dead": false,
"Pid": 0,
"ExitCode": 1,
"Error": "",
"StartedAt": "2017-03-22T00:51:59.234643501Z",
"FinishedAt": "2017-03-22T07:21:41.351927073Z"
},
"Config": {
...
"AttachStdin": false,
"AttachStdout": true,
"AttachStderr": true,
...
}
I could set docker to always restart, but I want to figure out what's the root cause of it. Any suggestions?

Answer 1

That happens to everyone. Lots of things can kill Express applications, like weird HTTP requests. The Docker log should show the exception. Add an uncaughtException handler to log the issue.
process.on('uncaughtException', (e) => {
console.error(e); // try console.log if that doesn't work
process.exit(10);
});
If you can't find an error in the Docker log, then instead of logging to the console, maybe you can log to a file (make sure it is in a volume that stays between Docker runs though).
People don't like admitting this but many applications actually just log the exception and eat it and keep going in the uncaughtException handler without exiting. Because things like broken requests or other stuff that doesn't matter often kill the server you can usually get away with this. But then once in a while you will get burned by something strange happening to the server state that it can't recover from and you will have no idea because you just ate the exception.
You may be able to have it autorestart the app https://docs.docker.com/docker-cloud/apps/autorestart/ which might be a good solution.
Otherwise look into an example of using pm2 along with Docker if possible, pm2 will handle restarting for you.