Requiring newrelic through NODE_OPTIONS results in out-of-memory exit

Description

I’m running a server with Next.js and want to bring newrelic in for monitoring. I’m not using a custom server with Next.js, so I don’t have a traditional entry-point file that I can put require('newrelic') at the top of. I can though require modules before my server starts, with NODE_OPTIONS="-r newrelic". This works locally for me and in some deployed environments.

However, in a memory limited environment, I’m experiencing intermittent out-of-memory kills (code 137). I suspected this was due to Next.js’ compatibility with newrelic, but creating a small, reproducible example with Next.js experiences the same behaviour.

Expected Behavior

The newrelic agent required with NODE_OPTIONS="-r newrelic" should give the same behaviour as require('newrelic').

Steps to Reproduce

I have two environment variables set, NEW_RELIC_APP_NAME and NEW_RELIC_LICENSE_KEY. To reproduce this, you’ll need to set them to valid values for your environment.

All commands start a Node alpine docker container, install newrelic via npm and attempt to run a “Hello world” node script. They are memory limited to 256MB - but this is arbitrary, as even 512MB gives similar results (it just takes longer to crash).

The below command runs require("newrelic") then console.log("Hello world"). The message will be logged, along with a warning about a missing newrelic.js config file - but the container will exit with code 0.

$ docker run --rm -w /tmp -m 256MB -e NEW_RELIC_APP_NAME -e NEW_RELIC_LICENSE_KEY node:16-alpine sh -c "npm install newrelic && node -e 'require(\"newrelic\");console.log(\"Hello world\")'"

The next command sets NODE_OPTIONS='-r newrelic then runs console.log("Hello world"). The container will exit with code 137, and intermittently this will happen before the message is logged (or the warning about a newrelic.js config is printed).

$ docker run --rm -w /tmp -m 256MB -e NEW_RELIC_APP_NAME -e NEW_RELIC_LICENSE_KEY node:16-alpine sh -c "npm install newrelic && NODE_OPTIONS='-r newrelic' node -e 'console.log(\"Hello world\")'"

This can also be reproduced via npm scripts (i.e. "start": "NODE_OPTIONS='-r newrelic' node ."), putting the commands into a shell script, or when running an npm command or node script.

I’ve reproduced it on node:16-alpine, node:14-alpine, node:16 and node:14.

Your Environment

• Browser name and version: N/A
• Node version: 16, 14
• Operating System and version: Ubuntu via WSL2, Alpine via Docker

Additional context

This could be unrelated to newrelic and is instead an error handling difference between require(...) and NODE_OPTIONS="-r ..." - but I wasn’t able to reproduce this with other packages.

[bizob2828]

@atkinchris Ok I traced it down in our codebase and logged https://github.com/newrelic/node-native-metrics/issues/155. A few options to get you around this issue if you do not want to wait for a fix:

• change from NODE_OPTIONS='-r newrelic' to node -r newrelic node_modules/.bin/next start

OR

• Disable native metrics module in config

    plugins: {
      native_metrics: {
        enabled: false
      }
    }

[atkinchris]

I’m no longer able to reproduce this bug after the update. Thanks for this! 🎉