Data guarantees on a Volume
Last updated 16 November 2026
A Volume is a disk of your account that one Machine mounts (see Volumes). This is what Tonbo promises about a write to it; what it promises about a database the Machine binds is on Data guarantees in a database.
What a write survives
fsync, below, is any flush:fdatasync,sync, a write opened withO_SYNCorO_DSYNC, or an unmount.closeon its own is not one.
Word for word
Your Volume lives on the disk of the server your Machine runs on. Tonbo copies it to regional storage about once a minute while the Machine is awake, and once more every time it sleeps.
- When
fsyncreturns, the write is on that server's disk. It survives a crash of your program, your Machine, or the Volume's software on that server.- It also survives the loss of that server or its zone once the next copy finishes, normally within a minute. If copying falls five minutes behind,
fsyncwaits until it catches up, so at most five minutes of writes are ever at risk.- If the server stops responding, Tonbo waits up to 10 minutes for it to come back, then restores the Volume elsewhere from its last finished copy.
- When a Volume is restored that way, it comes back exactly as it was at the last finished copy: all of it from one moment, never a mix.
- A sleeping Machine's Volume has finished its copy, so losing a server or a zone costs it nothing.
- Losing a whole region is not covered. A Volume is not a backup. If every
fsyncmust be safe off the server, use a Tonbo Database.
There are no Volume snapshots, backups or resizing, and a Volume does not protect against your own deletes.
Why a Volume keeps its data this way, and what that means for your program, is on How Volumes work; its limits, costs and rules are on Things to know.