Collecting…

Four checks, cheapest first: reachability and RTT, MTU and path MTU, each link alone, then all links at once. The last two are the decisive comparison — if the concurrent total is near a single link's throughput, the circuits share a bottleneck upstream and no software can work around it.

Every run is kept under /var/lib/ftm/runs/<run-id>/ with its summary, the settings in force, the shard split, every worker log and the throughput recorded while it ran. Click a row to review it.

Measures all three legs — network, source read, destination write — then sizes the settings around whichever is slowest. The destination is swept at several write-concurrency levels, because whether many parallel streams help or hurt depends entirely on the device at the far end.

Loading…

Pick which physical interfaces carry the transfer links. The management interface (the one this page arrived on) is never offered. Apply runs the same guarded netplan script as commissioning: it configures only these interfaces, checks the generated output leaves the management NIC alone, arms a dead-man revert, applies, and watches the management address for 45 seconds before it commits.

Pairing exchanges the transfer account's keys over the management network and records each unit's management address on the other. Both units must be running this UI and have opposite roles.

An attached disk, an NFS or SMB share, or an existing path. The result is mounted now, written to the destination unit's fstab so it comes back after a reboot, made writable by the transfer account, and recorded as DST_ROOT on both units.