Failing to refetch federated tracks

Steps to reproduce

1. Browse a track or album from a distant library, for which i (the current logged in user) have access
2. Play a track

What happens?

After a few seconds loading, the track doesn’t play. My browser get a 503 code for a /api/v1/listen/{{ track_id }} API call.

On the server logs, when I browse the album, i see a line with:

funkwhale_api.music.views INFO Refetching music.Album:{{ internal_id }} at {{ instance }}

followed, a few seconds after i try to play a track, by:

django.request ERROR Service Unavailable: /api/v1/listen/{{ track_id }}

Both log lines are generated by gunicorn. No celery log is written in the operation. I tried with LOGLEVEL=DEBUG in my env file, and while more logs are present, there’s nothing about a possible failure of the refetch task, nor a mere presence of this kind of task in the celery logs.

What is expected?

The track should play.

Context

Funkwhale version(s) affected: 1.0.1

I’m running Funwhale 1.0.1 on a ubuntu container, without docker (i’m using the ansible role). A few weeks ago, i migrated the whole server (i copied the entire database and media directory from the old server to the new). The problem could have started at the migration. Everything in my data directory is owned by the funkwhale user, that have the permission to write in the federation_cache subdir.

I tried to follow a new public library (on open.audio), and this time I could play the remote tracks.

I have the same problem when trying to play a track from a subsonic client (like dsub)