ENOENT: no such file or directory, scandir '.serverless/requirement'

[rsynnest]

When I run sls deploy I get the following error multiple times in a row: Error: ENOENT: no such file or directory, scandir '.serverless/requirements'

Each time, I verify that .serverless/requirements does in fact exist, it is a symlink to a cache folder of pip packages in my home directory. If I run sls deploy 3 or 4 times without changing anything, on the 3rd or 4th try it will run with no error. I am running on Win10/WSL2 but have never had issues with symlinks.

Serverless: Zipping required Python packages...

  Error --------------------------------------------------

  Error: ENOENT: no such file or directory, scandir '.serverless/requirements'

     For debugging logs, run again after setting the "SLS_DEBUG=*" environment variable.

  Get Support --------------------------------------------
     Docs:          docs.serverless.com
     Bugs:          github.com/serverless/serverless/issues
     Issues:        forum.serverless.com

  Your Environment Information ---------------------------
     Operating System:          linux
     Node Version:              12.18.1
     Framework Version:         1.78.1 (standalone)
     Plugin Version:            3.7.0
     SDK Version:               2.3.1
     Components Version:        2.33.4

[thekensman]

We're seeing this exact same issue with the following pythonRequirements section in our YML:

pythonRequirements:
    dockerizePip: non-linux
    zip: true
    #Get rid of unnecessary package files
    slim: true
    #But keep necessary binary files to avoid "ELF load command address/offset not properly aligned"
    strip: false

It fails and fails in Gitlab, but occasionally (every third try or so) it works. EDIT: The deploy also works consistently when I run it locally

[mapattacker]

I had the same exact issue... hope someone can help

[ghost]

Any update on this?

[TimekillerTK]

Issue

NOTE: Scroll down for workaround/solution.

Came across this issue in my Gitlab CI pipeline when running this command in a gitlab build job:

• sls package --region $AWS_REGION --stage $CI_ENVIRONMENT_NAME --package .serverless

And passing the created artifact to a gitlab deploy job which runs:

• sls deploy --region $AWS_REGION --stage $CI_ENVIRONMENT_NAME --package .serverless

Serverless Error ----------------------------------------
 
  Cannot read file artifact ".serverless/pythonRequirements.zip": ENOENT: no such file or directory, stat '.serverless/pythonRequirements.zip'
 
  Get Support --------------------------------------------
     Docs:          docs.serverless.com
     Bugs:          github.com/serverless/serverless/issues
     Issues:        forum.serverless.com
 
  Your Environment Information ---------------------------
     Operating System:          linux
     Node Version:              16.3.0
     Framework Version:         2.70.0 (local)
     Plugin Version:            5.5.3
     SDK Version:               4.3.0
     Components Version:        3.18.1

Why this happens

According to the serverless-python-requirements documentation, caching is enabled by default.

What this does is it changes the paths to (among other things) pythonRequirements.zip to a symlink. You can see this if you download the artifact to your local machine - in gitlab you can't see symlinks in the gui:

total 96
drwxr-xr-x@ 9 user  group    288 Jan  7 10:08 .
drwxr-xr-x@ 3 user  group     96 Jan  7 10:17 ..
-rw-r--r--@ 1 user  group   4986 Jan  7 10:08 project.zip
-rw-r--r--@ 1 user  group   2077 Jan  7 10:08 cloudformation-template-create-stack.json
-rw-r--r--@ 1 user  group  10379 Jan  7 10:08 cloudformation-template-update-stack.json
lrwxrwxrwx@ 1 user  group    133 Jan  7 10:08 pythonRequirements.zip -> /home/user/.cache/serverless-python-requirements/ff8b23b818b507958ef9b61047f665cde924fc7342f352dc2c6c1487333df041_x86_64_slspyc.zip
lrwxrwxrwx@ 1 user  group    129 Jan  7 10:08 requirements -> /home/user/.cache/serverless-python-requirements/ff8b23b818b507958ef9b61047f665cde924fc7342f352dc2c6c1487333df041_x86_64_slspyc
-rw-r--r--@ 1 user  group    131 Jan  7 10:08 requirements.txt
-rw-r--r--@ 1 user  group  20447 Jan  7 10:08 serverless-state.json

The result is that the symlinks of these files get passed on as artifacts to the next job instead of the files themselves. And since it's a different job, the .cache directory doesn't exist, which results in the file not found error.

Solution / Workaround

You can disable caching by setting both useDownloadCache & useStaticCache to false.

custom:
  pythonRequirements:
    useDownloadCache: false
    useStaticCache: false

This fixed it for me.

[Douglas019BR]

I did that and I still have this problem.

[rsynnest]

For what it's worth, in my original report of this issue I was running deploys manually on my local machine. There was no CI involved and it seemed to be purely an issue with the plugin.

[BwL1289]

I am also experiencing this

╰─ sls deploy                 
Running "serverless" from node_modules

Deploying ...omitted... to stage dev (us-east-1)
WarmUp: Creating warmer "default" to warm up 1 function

✖ Stack ...omitted... failed to deploy (9s)
Environment: darwin, node 14.17.4, framework 3.21.0 (local) 3.21.0v (global), plugin 6.2.2, SDK 4.3.2
Credentials: Local, "default" profile
Docs:        docs.serverless.com
Support:     forum.serverless.com
Bugs:        github.com/serverless/serverless/issues

Error:
[OperationalError: ENOENT: no such file or directory, scandir '.serverless/requirements'] {
  cause: [Error: ENOENT: no such file or directory, scandir '.serverless/requirements'] {
    errno: -2,
    code: 'ENOENT',
    syscall: 'scandir',
    path: '.serverless/requirements'
  },
  isOperational: true,
  errno: -2,
  code: 'ENOENT',
  syscall: 'scandir',
  path: '.serverless/requirements'
}

[BwL1289]

For anyone coming across this - it seems if slimPatternsAppendDefaults is set to false, this error occurs. After hours of trying, I turned slimPatternsAppendDefaults: true and commented out my slimPatterns section and was able to deploy.

Very frustrating.

[pavel-nesterov]

What helped me is adding requirements.txt to the service folder (next to the serverless.yml). I am deploying via Serverless'es CI/CD (by committing to the branch, not from local machine)

UPDATE: when solution above did not work - the only one worked was this setup in serverless.yml:

  pythonRequirements:
    layer: false
    usePipenv: false
    useStaticCache: false
    useDownloadCache: false
    slim: false