`jaco backup` and `jaco restore`

Cluster state export and restore. The backup tarball is a raft snapshot plus metadata; restore primes a fresh data directory so a single node can be the seed of a recovered cluster.

`jaco backup`

Synopsis

jaco backup --output <file>
            [--server <host:port> --token <op>]
            [--ca-cert <path>] [--socket <path>]

Flags

flag	default	meaning
`--server <addr>`	—	leader gRPC; omit to use the local socket
`--token <op>`	`JACO_TOKEN`	operator bearer token (with `--server`)
`--ca-cert <path>`	`/var/lib/jaco/node/ca.crt`	cluster CA PEM
`--socket <path>`	`/var/run/jaco/jaco.sock`	local jacod unix socket
`--output <file>`	— (required)	destination tarball (e.g. `cluster.tar.gz`)

Auth

Operator token (TCP) or unix-socket trust (local).

Streams a fresh raft snapshot plus a meta.json (cluster id, snapshot index/term, JACO version, taken-at timestamp, leader-at-snapshot) into a gzipped tar file at --output. The RPC has a 5-minute deadline; the CLI prints the total byte count on success.

The snapshot is consistent at a single raft commit index, so the restored cluster will reflect every deployment committed before that index and none committed after.

Exit codes

0 — backup written.
1 — auth, transport, or write error.

Examples

jaco backup --server $LEADER --output cluster-$(date +%F).tar.gz
# Wrote 41279 bytes to cluster-2026-05-25.tar.gz

`jaco restore`

Synopsis

sudo jaco restore --input <file> --name <hostname>

Flags

flag	default	meaning
`--input <file>`	— (required)	backup tarball
`--name <s>`	— (required)	hostname / raft local-id for this node

JACO_DATA_DIR overrides the target data directory (default /var/lib/jaco).

Auth

Filesystem only — restore writes to disk locally; it does not RPC the daemon. Run as root on the receiving host with jacod stopped.

Behavior

Primes the data directory from the backup: validates the metadata against the daemon's version, seeds a fresh raft store from snapshot.bin, and writes a marker so the daemon emits a RESTORE_COMPLETED audit event on its first boot.

After restore, start the daemon and confirm the cluster comes up as a single voter:

sudo systemctl start jaco
jaco cluster status

Additional nodes rejoin via the usual jaco node join flow.

Exit codes

0 — restored; next step is to start the daemon.
1 — bad input, version mismatch, or filesystem error.

Examples

sudo systemctl stop jaco
sudo jaco restore --input cluster-2026-05-25.tar.gz --name $(hostname)
sudo systemctl start jaco
jaco cluster status