How can I configure seaweedfs so that files under a certain directory explicitly hold files on a certain data center?

[nikp123]

To unwrap this question into something more readable (and hopefully understandable):

I have 3 machines, on every machine i have instances of the master, volume and filer services.

Each of these machines is in a different dataCenter.

I have configured the default replication of 100 so that the data has two copies scattered among 3 machines.

I soon ran into an issue where Jellyfin would be excruciatingly slow because it needs to read files on each scan if they're not locally available on the same machine.

Is there some sort of way to ensure that files written to a certain subdirectory will be stored on a given dataCenter.
I've tried using fs.configure but that doesn't seem to do anything.

A bit of guidance is needed, apologies if this sort of use case is not supported.

[Material-Scientist]

weed shell -master=xxx

fs.configure -locationPrefex=<directory_prefex> -dataCenter=<your_datacenter> -apply

Make sure you include the -apply.

Also:
The storage options are applied during write time.

So if a file is created and then moved to the configured path, the storage option would not take effect.

[nikp123]

This still doesn't seem to work, I've tried locationPrefixes with both the leading slash and without. Neither of those stores the files on the dataCenter it seems.

Because when I cat those files after they've been copied and pipe them through pv I see that the speed is limited to my internet access in spite of me specifying localhost as my filer.

[Material-Scientist]

1. "how can I be sure that the data landed on the appropriate target"
   You can either check the UI and see which volume servers the volumes are on, or you can check with the filer which volume your files are on.

2. "And will the write fail if it isn't able to write to the specified dataCenter"
   IIRC, the writes will still go through, since it's just a "preferred" setting.

3. Make sure you're not overriding with other configs in your volume, filer, and mount setups. They each allow you to specify the dc.

4. If you're using an embedded leveldb as filer store your changes may not be propagated immediately (eventually consistent). So, either wait, check filer logs to see if the config was reloaded, or use a shared filer store.

5. Make sure your volume servers & filer are on the right network interface (e.g. localhost/LAN instead of public IP). You can check in the UI or via weed shell > cluster.check

[nikp123]

Now I can definitely confirm that it still isn't respecting the suggestion to put files on the target dataCenter. I'll try later when I clear up a bit of space so it could allocate more volumes.

[nikp123]

Make sure you're not overriding with other configs in your volume, filer, and mount setups. They each allow you to specify the dc.

Can commandline flags I pass to the daemon's (volume and filer) interfere with this somehow. I currently only have my replication rules set there.

[nikp123]

I'm gonna mark this answer as correct for the sake of completeness, but in the end I ended up entirely breaking the filesystem :|

Don't worry though, the data is still fine as of now.