I wanted to share an opensource tool (Rust, AGPL v3) I’ve been working on called HuskHoard. If you’re tired of cold, untouched files consuming 3x raw block storage across your OSDs, this might help.
HuskHoard runs on your CephFS gateway node. It automatically backs up cold files to an S3 bucket or LTO Tape, and then issues a POSIX fallocate command. The CephFS kernel client natively translates this into RADOS object deletions—instantly freeing up the cluster space—while keeping the file's metadata completely intact on the MDS.
To the user, the file looks like it never left. If an application tries to open it months later, HuskHoard uses fanotify to pause the PID, fetches the payload back from cold storage, and recreates the RADOS objects seamlessly so the application doesn't crash.
If you are dealing with cluster expansion and want to offload data without breaking file paths, I’d love for you to check it out and let me know what you think! there is a technical articles here cold ceph
HuskHoard runs on your CephFS gateway node. It automatically backs up cold files to an S3 bucket or LTO Tape, and then issues a POSIX fallocate command. The CephFS kernel client natively translates this into RADOS object deletions—instantly freeing up the cluster space—while keeping the file's metadata completely intact on the MDS.
To the user, the file looks like it never left. If an application tries to open it months later, HuskHoard uses fanotify to pause the PID, fetches the payload back from cold storage, and recreates the RADOS objects seamlessly so the application doesn't crash.
If you are dealing with cluster expansion and want to offload data without breaking file paths, I’d love for you to check it out and let me know what you think! there is a technical articles here cold ceph