Using

Streams the archive with HTTP Range support for resumable downloads.

applies to v0.5.1

GET

/store-api/backups/{id}/download

const url = 'https://example.com/store-api/backups/example/download';
const options = {method: 'GET', headers: {Authorization: 'Bearer <token>'}};

try {
  const response = await fetch(url, options);
  const data = await response.json();
  console.log(data);
} catch (error) {
  console.error(error);
}

curl --request GET \
  --url https://example.com/store-api/backups/example/download \
  --header 'Authorization: Bearer <token>'

The archive is NDJSON. Its first line is the manifest, format version 2: format_version, id, created_at, luradb_version, scope, include_auth, encoding, schedule, and

shards: data shards of the server that wrote it. Informational only: a restore places every domain by the shard count of the server it runs on, so an archive restores into any shard count.
lsn_cut: {kv, json, rel}, per engine a write-ahead-log position that every write in the archive is at or below (a lower bound of the cut); 0 for an engine outside the scope. Positions keep growing across restarts, so they compare across runs of the server.
kv_snapshot_ts, json_snapshot_ts, rel_snapshot_ts: per engine, the newest read-snapshot stamp among the data shards; 0 for an engine outside the scope.

All shards of one engine are read at one consistent cut: an archive that holds a write holds every write confirmed before it. The relational cut is taken before the KV and JSON cuts. Domain lines carry their domain’s name; sections follow in the order KV, JSON, relational, auth; the last line is the checksum.

A relational domain is a rel-domain line (name, created_at), then its tables by creation, its views, and per table its rows and AUTOINCREMENT state:

rel-table: domain, name, columns (each name, type as SQL type name, nullable, primary_key, autoincrement, unique, an optional default of {"kind": "null"}, {"kind": "current_timestamp"} or {"kind": "literal", "value": …}, an optional references naming the target table) and indexes (name, column, unique; every index under its stored name).
rel-view: domain, name, sql, the stored SELECT text.
rel-row: domain, table, row, every column by name, NULL as null, KVREF/JSONREF links as stored.
rel-seq: domain, table, hw: the next insert without a key gets hw + 1.

Values are JSON numbers (INTEGER, REAL), booleans, strings (TEXT, KVREF, JSONREF) and, for TIMESTAMP, integer milliseconds since the Unix epoch. Archives of format version 1, which lack shards and lsn_cut, and archives without relational lines remain restorable.

Authorizations

bearer_auth

Parameters

Path Parameters

required

string

Backup id

Responses

200

Full NDJSON archive

application/x-ndjson

206

Partial content (Range request)

application/x-ndjson

400

Invalid backup id

text/plain

string

404

Backup not found

text/plain

string

409

The backup job for this id is still running

text/plain

string

503

Backup is disabled

text/plain

string