Closed AgDude closed 3 years ago
This is already happening at https://github.com/danielfrg/s3contents/blob/d7e398c7da8836ac7579fa475bded06838e053ea/s3contents/s3_fs.py#L91 in version 0.12.1
And when I do jupyter notebook --log-level DEBUG
I can see:
[I 21:31:55.874 NotebookApp] [nb_conda_kernels] enabled, 4 kernels found
[D 21:31:55.929 NotebookApp] S3contents.S3FS: Making dir: `mybucket.emptytemp2/notebooks/.s3keep`
[D 21:31:57.539 NotebookApp] S3contents.S3FS: Listing directory: `mybucket.emptytemp2/notebooks`
[D 21:31:58.561 NotebookApp] S3contents.S3FS: `mybucket.emptytemp2/notebooks` is a directory: True
So, this is already happening. You need to ensure that your prefix has no starting / and ending /
c.S3ContentsManager.bucket = 'mybucket.emptytemp2'
c.S3ContentsManager.prefix = 'notebooks'
I set up a new deployment with a bucket and prefix. Without a normal file in s3 with this prefix, jupyter fails to start with the following error:
FileNotFoundError: my-bucket/my/prefix
This is related to this issue in s3fs: https://github.com/dask/s3fs/issues/170
It think the right solution is to catch this and drop a file into s3. I uploaded a text file through the AWS console, which allowed jupyter to start.