Skip to content

Snapshots

Updated

A snapshot captures a running machine so you can put it back exactly as it was. Snapshots are available for classic VMs.

One snapshot is always both the disk and the memory, taken together in a single moment. The machine pauses briefly while the memory is written out, then resumes on its own.

The two halves are never separated, and that is a correctness requirement rather than a convenience. Restoring memory alone would put the guest’s view of its filesystem back in time while the disk stayed in the future, and the next writes would corrupt it silently.

Open a running machine, go to its snapshots, and give the snapshot a name of up to 64 characters. Expect a short pause, roughly the time it takes to write the guest’s memory to disk.

Plan for the space: each snapshot stores about as much as the guest’s memory size, plus disk growth as the machine keeps writing afterwards.

Snapshots fail loudly and early rather than producing something that cannot be restored safely.

Refused whenWhy
The machine is not runningThe point of the snapshot is the running memory state
A data volume is attachedThe volume would not roll back with the memory, leaving it stranded in the future. Detach it first
A PCI device is passed throughThe device’s state lives outside the machine and cannot be captured
It is an AppVMAppVMs roll back by redeploying instead. See Logs, exec, and redeploy
The machine’s shape changedRestore requires the same vCPU and memory shape as when the snapshot was taken

Each machine keeps up to eight snapshots. At the limit you must delete one yourself. Virtainer Lite will not silently evict your oldest restore point to make room for a new one.