Move sandboxes between hosts
To move a sandbox’s saved state from one of your machines to another, make a named snapshot, export it as an archive, copy the archive, then import and resume on the other host. This is not live migration: it does not transfer an active connection or keep the source and destination in lockstep.
Snapshot → export → copy → import → resume
Section titled “Snapshot → export → copy → import → resume”On the source:
bhatti snapshot create dev --name dev-transferbhatti snapshot export dev-transfer --include-base -o dev-transfer.tar.zstscp dev-transfer.tar.zst other-host:~/On the destination (with its own bhatti server and API credentials):
bhatti snapshot import ~/dev-transfer.tar.zstbhatti snapshot resume dev-transfer --name devYou can stream instead of keeping an intermediate file: bhatti snapshot export dev-transfer --include-base | ssh other-host 'bhatti snapshot import -'. -o is --output; without it, export writes binary to stdout and refuses a terminal. Import accepts a path or - for stdin, and --name if the saved name conflicts with another snapshot. The archive is tar + zstd. After import, it remains a named snapshot; only resume creates a sandbox. Exporting never destroys or stops the source.
The destination must have the same architecture as the source: arm64 and x86_64 snapshots cannot cross. For a default memory snapshot, it must also use the same host OS and a CPU with the features used by the saved guest. A Linux KVM memory snapshot cannot resume on macOS HVF, even if both hosts are arm64. --type filesystem snapshots capture only the root disk: they can move Mac↔Linux on the same architecture and cold-boot on the destination, with no running processes preserved. CPU checkpoint feature checks apply to memory snapshots. Import refuses incompatible archives.
For a same-architecture move between a Mac and a Linux host, take a filesystem snapshot rather than the default memory type. Preserve attached volume data separately; this type does not retain volumes or the source’s network policy:
bhatti snapshot create dev --name dev-disk --type filesystembhatti snapshot export dev-disk --include-base -o dev-disk.tar.zst# Copy the archive, then on the other OS:bhatti snapshot import dev-disk.tar.zstbhatti snapshot resume dev-disk --name devThe root disk is a copy-on-write overlay depending on an immutable content-addressed base. Without --include-base, import looks for identical base content already on the destination; the same tier name alone is not enough. Use --include-base when in doubt, at the cost of a larger archive. A memory-snapshot archive includes RAM/CPU state, root disk changes, attached-volume copies and network policy; resumed volumes are independent copies, re-attached at their saved mount points. A filesystem-snapshot archive carries only the root disk and manifest, not attached volumes or the source’s network policy. Export refuses snapshots with --mount host-directory binds, or attached qcow2 volumes that depend on external backing images. It does not migrate server-side secret records or grant authorizations — provision those separately. Credentials already present in captured guest RAM or files remain included, so protect the archive as sensitive. Credentials explains the boundary.
For repeated local experiments, fork a sandbox without exporting. For filesystem durability before moving a checkpoint, see Data & durability.