Tommie Gannert

This is a research project, and is not meant for production use.

The idea is to have a central archive of backup snapshots for workstations. Each workstation uploads to a new snapshot, using the rsync --hard-link feature to use the previous snapshot as a base. Periodically a job is run to garbage collect old snapshots and to ensure hard links are used consistently for all hosts.

The archive is divided into trees. Each tree is a shared folder. When multiple hosts use the same tree, they upload to separate snapshot directories and a periodic job merges the latest snapshot directories to form download snapshots. Each host then downloads its download snapshot to get files available on other hosts. Collisions are managed by a pluggable strategy that takes the latest version of the file from all hosts, and the hostname to generate the download snapshot for. It produces a set of new files in the snapshot directory. Further improvement is using a union file system to handle creation of download snapshots.

One example of a strategy is simply taking the latest file (by snapshot timestamp or mtime). The other files are made available through a web interface. Another example is adding a host suffix to all files but the host's own and let the host download all the files.