Add ability to map uploads folder

[stuartshields]

Add ability to map uploads folder

[roborourke]

Yeah I had wanted to use minio to provide this in the past as it has a file sync feature. The problem was that fake S3 doesn’t use a regular file structure for storage (more of a database) so it didn’t work to just mount the volume at the time.

If it’s that simple and works then great, we should definitely do this!

[roborourke]

Can you show me what you did? I just tried this and it's as I suspected the real file structure is like this:

└── uploads
    ├── 2020
    │   └── 03
    │       └── Screenshot-2020-03-04-16.37.29.png
    │           └── .fakes3_metadataFFF
    │               ├── content
    │               └── metadata

The content file in the Screenshot-2020-03-04-16.37.29.png directory is a binary blob and as such can't be opened properly locally without modifying the file name.

I think we'd need to look at switching to minio and using the sync feature to keep the uploads directory in sync. This is what I had to do for the chassis-minio extension when I needed to work on S3-Uploads previously.

[stuartshields]

@roborourke I ran into the same issue, because time is critical I went back to using Chassis instead. Your idea for minio sounds like the best way forward.

[roborourke]

In terms of adding assets it could perhaps be done using the s3 cli commands. Will see if it's something we can at least document in the interim.

[goldenapples]

@stuartshields Were you looking at the same issue here I was in #142? If so, the CLI command that ended up getting me unstuck might also be of use for you...

[roborourke]

@goldenapples @stuartshields I've opened #145 which sort of solves this. I was hoping to get bi-directional sync working with minio (alternative to fakes3) however it was a bit fraught. What that PR does do however is the following:

• Syncs S3 to the host's content/uploads directory automatically and continuously
• Adds a composer server import-uploads command to sync files from /content/uploads into S3

Does that seem satisfactory as a solution? I think we'll still need to write some guides for importing existing data and uploads into both environments regardless.