Backup and storage

A tie deployment has two data sets:

• the triplestore — one .tie file per collection (plus _prev history);
• the filehost — the content-addressed blob directory.

They can be backed up independently, but restore them together.

The simplest backup

# stop writes, or use a snapshot
tie dump --file triples.tsv
rsync -a ~/.local/share/tie/ backup/tie-db/
rsync -a /var/lib/tie/data/ backup/tie-data/

tie dump is a TSV of every triple; tie restore is additive and idempotent. The blob store is content-addressed, so a plain copy is valid: a blob is either present or not, and its name is its hash.

Filesystem snapshots

On Btrfs, a read-only snapshot of the triplestore directory and the blob directory is consistent enough for btrfs send:

btrfs subvolume snapshot -r /var/lib/tie /var/lib/tie-snap
btrfs send /var/lib/tie-snap | ssh backup btrfs receive /backup

contrib/backup/ has btrfs-send-backup.sh, rsync-backup.sh and triplestore-db-backup.sh.

The reaper

The filehost can reap blobs that are no longer referenced. ReapInterval = "0" disables it (the default in the test environment). If you enable it, make sure the triplestore backup you keep still references what the blobs need.

Restore order

1. Stop the services.
2. Restore the blob directory.
3. Restore the .tie file(s).
4. Start the triplestore, then the filehost.
5. tie verify --deep and tie verify --check-blobs.

The source note is docs/filehost-storage-and-backup.md.