Unable to use Cloudflare R2's presigned url's for media transcoding output destination

[uncvrd]

Hi there! I've been trying to implement audio transcoding (wav -> mp3) using the transcode endpoint, but I receive the following error:

@foundrydev/trpc-web:dev: {
@foundrydev/trpc-web:dev:   path: '/media/transcode',
@foundrydev/trpc-web:dev:   status: 'Failed',
@foundrydev/trpc-web:dev:   progress: 1,
@foundrydev/trpc-web:dev:   api_version: 'v1.7',
@foundrydev/trpc-web:dev:   error: {
@foundrydev/trpc-web:dev:     type: '/problems/output-file-access',
@foundrydev/trpc-web:dev:     title: 'Output location access problem',
@foundrydev/trpc-web:dev:     detail: 'Failed to generate output media due to concurrent writes to the same output location.'
@foundrydev/trpc-web:dev:   }
@foundrydev/trpc-web:dev: }

Essentially it gets stuck at 90% progress for a few minutes before showing this error. I've been able to successfully use AWS S3 and Digital Ocean Spaces to output to my destination directory but for some reason R2 has issues with however Dolby is uploading the audio files to the bucket. Was curious if anyone could provide insight on how the api is confirming that the file is uploaded?

The docs say for this error:

output-file-access

This is a write error if the URL provided as output is inaccessible. The endpoint must accept a PUT request along with a concluding HEAD or GET request that validates the full file has been stored.

Which sorta conflicts with what the error logs say. One says it errors since there are concurrent writes while the other claims that this url must be able to provide HEAD/GET to validate (which I'm not sure how that's possible with a put presigned url 🤔 )

Here's how I have implemented, nothing too special! I use NodeJS.

const command = new PutObjectCommand({
    Bucket: this._bucket,
    Key: key,
});

const url = await getSignedUrl(this._s3Client, command, {
        expiresIn: 3600,
});

const { data } = await this.post("transcode", {
    data: {
        inputs: [
            {
                source: "MY SOURCE PRESIGNED URL",
            },
        ],
        outputs: [
            {
                kind: "mp3",
                destination: url,
            },
        ],
    },
});

I configure my credentials for the NodeJS client like this

config: {
    credentials: {
        accessKeyId: env.R2_ACCESS_KEY,
        secretAccessKey: env.R2_SECRET_ACCESS_KEY,
    },
    region: env.R2_REGION,
    endpoint: env.R2_ENDPOINT,
},
bucket: env.R2_BUCKET,

Just as a reminder, I got a different S3 compatible library (Digital Ocean Spaces) working with this method, so I'm wondering what's up with R2 that just doesn't work the same way.

Thanks a lot for your time and any assistance here. It's really appreciated.

EDIT: I should note that the file is actually written to my bucket but for some reason the API is attempting to confirm that the file has been written but cannot do so?

[j12y]

Each cloud storage provider does behave a little bit differently with respect to being able to verify success with storing a file using a pre-signed URL. Since this isn't one of the already documented service providers (S3, GCP Cloud Storage, and Azure Blob Storage), it might require a bit more investigation and/or involvement from the team to investigate.

I'd suggest for this case, it might be worth submitting a support ticket if using Cloudflare R2 is a requirement for your project.

https://support.dolby.io/

[uncvrd]

Thanks Jayson! Left a support ticket