Where Does JuiceFS Store Metadata and Chunks?
Where self hosted JuiceFS keeps metadata and chunks on Linux servers. Exact folders are covered below.
Last updated
JuiceFS splits metadata from bytes with server storage. The metadata engine holds names and layouts. Object buckets hold content chunks. The release version matters for restores.
Same metadata plus objects means same file system. Exact buckets vary by S3 versus local flags. Keep dumps of metadata beside bucket snapshots. One without the other restores bytes without names. For moves, freeze writes first to avoid split brains. Check the in app storage readout when sizes look wrong on different releases.
Where JuiceFS stores this, by platform
/var/lib/juicefs (meta dumps plus buckets per flags)
Names live in the metadata engine with bytes in object buckets. Back up both together stopped. Client holds cache only. Buckets follow your storage flags.
Frequently asked questions
How do I migrate JuiceFS to a new server?
Back up the metadata engine dump plus object storage buckets with services stopped, then install the same release on the new servers before restore. Files follow chunks. Mixed releases can refuse old formats, so match versions first.
Where are JuiceFS chunks stored?
In object storage buckets with the metadata engine tracking names. The client holds login plus cache only. Sync metadata dumps regularly, since buckets alone restore bytes without names. Treat metadata plus objects as one backup unit.
Notice an outdated path? Let us know.