Non-deterministic CLI behavior

[drewbo]

Describe the Bug

Running CLI commands produces different results at different times with no other system changes. It's possible this is due to specific system constraints but no errors are produced. Example:

❯ pnpm payload migrate:status

> [email protected] payload
> cross-env NODE_OPTIONS=--no-deprecation payload migrate:status

<no output>
<wait 30 seconds>

❯ pnpm run payload migrate:status

> [email protected] payload
> cross-env NODE_OPTIONS=--no-deprecation payload migrate:status

[11:15:33] WARN: No email adapter provided. Email will be written to console. More info at https://payloadcms.com/docs/email/overview.
[11:15:33] INFO: Reading migration files from <redacted>/pages-editor/src/migrations
┌─────────────────┬───────┬─────┐
│            Name │ Batch │ Ran │
├─────────────────┼───────┼─────┤
│ 20250210_142843 │       │  No │
└─────────────────┴───────┴─────┘
[11:15:33] INFO: Done.

Link to the code that reproduces this issue

https://github.com/cloud-gov/pages-editor/

Reproduction Steps

This happened running a variety of migration commands as shown above. The database is PostGRES running locally

Which area(s) are affected? (Select all that apply)

area: core

Environment Info

Binaries:
  Node: 20.18.1
  npm: 10.8.2
  Yarn: 1.22.22
  pnpm: 10.2.1
Relevant Packages:
  payload: 3.20.0
  next: 15.1.6
  @payloadcms/db-postgres: 3.20.0
  @payloadcms/email-nodemailer: 3.20.0
  @payloadcms/graphql: 3.20.0
  @payloadcms/live-preview: 3.20.0
  @payloadcms/live-preview-react: 3.20.0
  @payloadcms/next/utilities: 3.20.0
  @payloadcms/payload-cloud: 3.20.0
  @payloadcms/plugin-form-builder: 3.20.0
  @payloadcms/plugin-nested-docs: 3.20.0
  @payloadcms/plugin-redirects: 3.20.0
  @payloadcms/plugin-search: 3.20.0
  @payloadcms/plugin-seo: 3.20.0
  @payloadcms/richtext-lexical: 3.20.0
  @payloadcms/translations: 3.20.0
  @payloadcms/ui/shared: 3.20.0
  react: 19.0.0-rc-66855b96-20241106
  react-dom: 19.0.0-rc-66855b96-20241106
Operating System:
  Platform: darwin
  Arch: arm64
  Version: Darwin Kernel Version 23.6.0: Fri Nov 15 15:13:56 PST 2024; root:xnu-10063.141.1.702.7~1/RELEASE_ARM64_T8103
  Available memory (MB): 16384
  Available CPU cores: 8

[jmikrut]

Hey @drewbo — we will look into this immediately. Thanks for the heads-up!

[jmikrut]

An update - this is caused by your use and import from @lexical/markdown - see this PR for a change we had to make because of that package:

#9382

The long and short is that for some reason, that package breaks tsx transpilation and Node fails silently. It looks to be because of prism and some type of circular dependency scenario with async imports. Low-level node stuff.

@AlessioGr will know more here but here's what you can do for now:

Change this import:

import { $convertToMarkdownString, TRANSFORMERS } from '@lexical/markdown';

To this:

import { $convertToMarkdownString, TRANSFORMERS } from '@payloadcms/richtext-lexical/lexical/markdown';

I have confirmed that this resolves the issue in your project.

We re-export these utilities for this exact reason. Would love for Lexical to look into this themselves (we've reported it) so others won't fall into this insane "gotcha" but for now, this is a workaround. Will mark this as a discussion so others can benefit from this!

[drewbo]

Thanks for the quick response @jmikrut. @AlessioGr Not sure if this is helpful but I got both a successful and unsuccessful run under my old configuration while running NODE_DEBUG=cluster,net,http,fs,tls,module,timers payload migrate:status to see some additional logging.

...
MODULE 2152: load "/home/vcap/app/node_modules/prismjs/components/prism-swift.js" for module "/home/vcap/app/node_modules/prismjs/components/prism-swift.js"
MODULE 2152: looking for "/home/vcap/app/node_modules/prismjs/components/prism-typescript.js" in ["/home/vcap/app/node_modules","/home/vcap/app/.node_modules","/home/vcap/app/.node_libraries","/home/vcap/deps/0/node/lib/node"]
MODULE 2152: load "/home/vcap/app/node_modules/prismjs/components/prism-typescript.js" for module "/home/vcap/app/node_modules/prismjs/components/prism-typescript.js"
MODULE 2152: looking for "/home/vcap/app/node_modules/prismjs/components/prism-java.js" in ["/home/vcap/app/node_modules","/home/vcap/app/.node_modules","/home/vcap/app/.node_libraries","/home/vcap/deps/0/node/lib/node"]
MODULE 2152: load "/home/vcap/app/node_modules/prismjs/components/prism-java.js" for module "/home/vcap/app/node_modules/prismjs/components/prism-java.js"
MODULE 2152: looking for "/home/vcap/app/node_modules/prismjs/components/prism-cpp.js" in ["/home/vcap/app/node_modules","/home/vcap/app/.node_modules","/home/vcap/app/.node_libraries","/home/vcap/deps/0/node/lib/node"]
MODULE 2152: load "/home/vcap/app/node_modules/prismjs/components/prism-cpp.js" for module "/home/vcap/app/node_modules/prismjs/components/prism-cpp.js"

<the prior logs are always printed, successful logs continue here>

MODULE 2152: Module._load REQUEST net parent: /home/vcap/app/node_modules/pg/lib/stream.js
MODULE 2152: load built-in module net
NET 2152: pipe false undefined
NET 2152: connect: find host <redacted db host>
NET 2152: connect: dns options { family: undefined, hints: 32 }
NET 2152: connect: autodetecting
...

It looks like it either hangs on the final prism-cpp.js transpilation or fails to start the pg/lib/stream.js transpilation. I didn't see anything obvious about potential timeouts in either case