Re-deploy my nodejs app and sadly got crashed.

[Mal]

I did this yesterday, it’s still okay, I don’t know why it crashed today in this Error
" Error: EACCES: permission denied, mkdir ‘/layers/heroku_nodejs-npm-install’"
and this is my fly.toml:

[Mal]

Could someone help me that?

[Mal]

I’ve realized this issue is not caused by this platform, while issue exists on image,
Fianlly, I decide to make an image by myself as demonstrated below:

and remove my build section of fly.toml
It’s also worked.

[asgeo1]

I’ve just started getting this exact same issue. In my fly.toml

[build]
  builder = "heroku/buildpacks:20"

[build.args]
  PORT = "8080"
  NODE_ENV = "production"

This was working fine 2 weeks ago, and now it no longer works. It will build and deploy, but the machine/app won’t start up, getting same start up errors per your screenshot.

I feel like something has changed with the buildpacks?

[davedx]

I have the same issue.

  Configuring firecracker
  [    0.048879] PCI: Fatal: No config space access function found
   INFO Starting init (commit: 913ad9c)...
   INFO Preparing to run: `launcher npm run db:migrate:latest:prod` as heroku
   INFO [fly api proxy] listening at /.fly/api
  2024/03/04 10:56:23 listening on [fdaa:0:be54:a7b:fd:6ea2:a139:2]:22 (DNS: [fdaa::3]:53)
  Machine created and started in 21.074s
  > controller@1.0.0 db:migrate:latest:prod
  > knex --knexfile ./knexfile.cjs migrate:latest
  Using environment: production
  npm ERR! code EACCES
  npm ERR! syscall mkdir
  npm ERR! path /layers/heroku_nodejs-npm-install
  npm ERR! errno -13
  npm ERR! Error: EACCES: permission denied, mkdir '/layers/heroku_nodejs-npm-install'
  npm ERR!  [Error: EACCES: permission denied, mkdir '/layers/heroku_nodejs-npm-install'] {
  npm ERR!   errno: -13,
  npm ERR!   code: 'EACCES',
  npm ERR!   syscall: 'mkdir',
  npm ERR!   path: '/layers/heroku_nodejs-npm-install'
  npm ERR! }

I really don’t want to maintain my own dockerfile. It feels like this is one of the reasons to use Fly, so I don’t have to mess about with that. Why can’t I just deploy like I could a few days ago?

[ilovetheory]

I am getting this same issue, my app has been working fine and suddenly started getting this error.

[loldedmded]

Anyone found a workaround on this yet? I am facing the same issue.

[lubien]

Hi @loldedmded

Are you using a buildpack? We created an open source project that during fly launch creates a custom Dockerfile without you having to worry about anything.

In case you app already exists try running npx --yes @flydotio/dockerfile@latest and removing the buildpack from your fly.toml from [build] and try deploying it again.

Here’s all options and source code for our Dockerfile generator GitHub - fly-apps/dockerfile-node: Dockerfile generator for Node.js

[a4z]

arg, seems I also got hit by that

Deploy crashes, for no to me unsterstandable reasons, but nothing chagned

I did a minor update in some JS script, deployed the app, and it’s not starting I do not understand what it going on, the app is running locally what are those npm errors in the log, are they the reasons, and how can I fix them? please help, my site is off now, since the last deploy seems to be gone … 2024-03-10T22:09:30Z proxy[e28650e0f67458] fra [info]Starting machine 2024-03-10T22:09:30Z app[e28650e0f67458] fra [info][ 0.041777] PCI: Fatal: No config space access function found 2024-03…

how can I fix that,

I do not fully understand that:

in case you app already exists try running npx --yes @flydotio/dockerfile@latest and removing the buildpack from your fly.toml from [build] and try deploying it again.

[littletheorem]

So I, too, was part of the Great Heroku Migration in late 2022 and only ran into these issues today (after ignoring the deprecation warnings for months, I think). Before I go into the step-by-step of how this worked for me, I want to say being able to roll back to the latest working image was very comforting, so I’ll mention that first.

First I grabbed the latest working image using this CLI: flyctl releases --image

I got the image ID from the right hand side, it will look like registry.fly.io/<your_app>@sha256:475309457929234782760586795467920762

Then I used this command to roll back to that image:
flyctl deploy -i registry.fly.io/<my_app>@sha256:<longUID>

To move forward with a NEW deployment, this worked for me (getting the exact same error as above).

1. Navigate to the file folder containing your fly.toml. I have another file here called Dockerfile.
2. Copy the contents of Dockerfile over to somewhere else (just to be safe).
3. Run npx --yes @flydotio/dockerfile@latest in this directory from the CLI.
4. It will run into the issue of Dockerfile already existing. It will ask for permission to overwrite it. Type y and hit enter.
5. Once it finishes, go to your fly.toml and remove the line under [build].

So this:

[build]
  builder = "heroku/buildpacks:20"

will become this (yes I left a line intentionally blank!):

[build]

6. Then deploy your app (for me this is flyctl deploy -a <my_app>)

My app is pretty simple, so this worked for me. I hope this helps someone!

[rubys]

FYI: if you rename your new Dockerfile to something like Dockerfile.fly, you can put the original Dockerfile back and tell fly which one to use by setting the name in your fly.toml:

Fly Launch configuration (fly.toml)

[lekanbar]

Thank you, I was already losing it. These steps worked perfectly for me as well!