NoSuchKey, Key not found

[htetlynnhtun]

I'm trying to setup b2 private bucket with cloudflare following this guide. https://www.backblaze.com/docs/cloud-storage-deliver-private-backblaze-b2-content-through-cloudflare-cdn. But it's not working. Every request to the contents is returning 404 not found

<?xml version="1.0" encoding="UTF-8" standalone="yes"?>
<Error>
    <Code>NoSuchKey</Code>
    <Message>Key not found</Message>
</Error>

[metadaddy]

Hi @htetlynnhtun - apologies for the delay in replying. Did you get this working? How did you configure the worker?

[htetlynnhtun]

Can't get it working for private bucket.
I added these vars

B2_APPLICATION_KEY_ID = ""
B2_ENDPOINT = ""
BUCKET_NAME = ""
ALLOW_LIST_BUCKET = ""

in the wrangler.toml file. And configure the B2_APPLICATION_KEY in the cloudflare dashboard as a secret.

[metadaddy]

Hi @htetlynnhtun - that XML response indicates that the request is reaching B2, but, somehow, the filename is not correct. Are you using this with rclone's --b2-download-url option? If so, you need to set the RCLONE_DOWNLOAD variable to "true".

[toddke]

I'm getting the same errors. My configuration is similar to @htetlynnhtun.

I'm not using rclone; instead, I'm hitting the endpoint using the browser. If I attempt to view the top-level directory [eg. https://b2.foo.bar/bucket-name], I get a response containing a recursive list of all files / folders within the bucket [I have set ALLOW_LIST_BUCKET = "true"]. However, as soon as I attempt to view a sub-directory [eg. https://b2.foo.bar/bucket-name/directory/], I receive the NoSuchKey response.

As a side note, I can download individual files. I get NoSuchKey only when viewing directories [which is critical for my use case].

[ruich-4]

Hi @toddke

I'm guessing you're setting BUCKET_NAME to $path.

Here's why you get "NoSuchKey" when accessing directories in your B2 setup:

function isListBucketRequest(env, path) {
    const pathSegments = path.split('/');
    return (env['BUCKET_NAME'] === "$path" && pathSegments.length < 2) || 
           (env['BUCKET_NAME'] !== "$path" && path.length === 0);
}

This code only allows listing for:

• Top-level bucket requests (https://b2.foo.bar/bucket-name)
• Empty path requests (https://bucket-name.b2.foo.bar/)

For subdirectories like /bucket-name/directory/:

• S3/B2 doesn't have real directories - they're just key prefixes
• The request looks for an object literally named "directory/"
• Since this object doesn't exist, you get NoSuchKey

That's why you can:

• List top-level directory
• Download individual files
  But can't browse subdirectories.

[toddke]

@ruich-4 Thanks for the great explanation. After digging into S3 I expected this was the case.

Since the S3 objects are created by B2, each "directory" could still have a generated object / key.

However, this is probably a niche use case and not likely worth the time / effort to implement.

[metadaddy]

Thinking about this, it might be possible to list the objects under a given prefix ('directory') when the client attempts to get that URL. I'll have a think about it.