Error: write EPIPE after upgrade to vr 9

[kvnam]

Hey,

I upgraded serverless-offline to 9.1.4 from 8.3.1 and am unable to run my Serverless app, which is a Graphql Express/Node backend. Any request to it crashes immediately with this error.

The error

Error: write EPIPE
    at ChildProcess.target._send (node:internal/child_process:866:20)
    at ChildProcess.target.send (node:internal/child_process:739:19)
    at ChildProcessRunner.run (file:///<my_path>/node_modules/serverless-offline/src/lambda/handler-runner/child-process-runner/ChildProcessRunner.js:57:18)
    at HandlerRunner.run (file:///<my_path>/node_modules/serverless-offline/src/lambda/handler-runner/HandlerRunner.js:134:25)
    at async LambdaFunction.runHandler (file:///<my_path>/node_modules/serverless-offline/src/lambda/LambdaFunction.js:289:20)
    at async file:///<my_path>/node_modules/serverless-offline/src/events/http/HttpServer.js:556:18
    at async exports.Manager.execute (<my_path>/node_modules/@hapi/hapi/lib/toolkit.js:60:28)
    at async Object.internals.handler (<my_path>/node_modules/@hapi/hapi/lib/handler.js:46:20)
    at async exports.execute (<my_path>/node_modules/@hapi/hapi/lib/handler.js:31:20)
    at async Request._lifecycle (<my_path>/node_modules/@hapi/hapi/lib/request.js:371:32)
    at async Request._execute (<my_path>/node_modules/@hapi/hapi/lib/request.js:281:9)

Environment: darwin, node 16.16.0, framework 3.21.0 (local), plugin 6.2.2, SDK 4.3.2

Serverless is at "^3.20.0"

When I downgrade to 8.8.1, it works fine. I have read the release notes, and my serverless and node are both the required versions. I know a change was made with regards to worker threads, but I'm unsure of what I need to do here about that. Any pointers on how to resolve this would be great. Thanks!

[dnalborczyk]

thank you for the report @kvnam .

just checked the code, I introduced a bug in v9.1.1, v9.1.0 should still work for you.

it seems that you are using child processes, with the flag --useChildProcesses (or as an option in your serverless.yml). if you are unsure why you are using that flag I would suggest to remove it and give worker threads (the default now) a try. if you want auto-reload of the handlers for development you can use the flag --reloadHandler, which reloads handlers on each request automatically.

Regardless, I should have a fix for the child process problem ready soon.

[kvnam]

Thanks @dnalborczyk! You are right I did have useChildProcesses set in the serverless.yml, missed it completely. I'll switch to 9.1.5 with your fix.