Crate backup 

Continuous backup of a mushroomdb store to S3-compatible object storage.

A BackupWorker runs beside the server. Every wal_interval_ms it ships the WAL bytes fsynced since the last tick as one immutable segment, {prefix}/wal/{generation}/{start:020}-{end:020}; when a snapshot is written it uploads snapshot.bin, any new wal.N.archive and the side files; and it writes manifests ({prefix}/manifest/{commit_seq:020}.json, pointed to by {prefix}/LATEST) naming consistent restore points.

The recovery point is the segment cadence, not the snapshot interval: a commit is off-box durable once the segment holding it is uploaded.

Before it uploads anything the worker takes the prefix’s BucketLease, an advisory single-writer lock, and while another instance holds a fresh one it uploads nothing and reports itself unhealthy (see lease).

restore_from_s3 reads a restore point back — the newest, or the one holding a point in time — verifies every object, and installs it as a store with a new backup lineage (see restore). boot::restore_on_boot does the same into an empty data directory when a server starts, and hands the bucket lease it took to the worker (see boot).

The worker never takes an engine lock and never writes to the store; a failed upload is retried with back-off and reported through BackupHandle, never to the write path.

Re-exports

pub use boot::restore_on_boot;

pub use boot::BootError;

pub use boot::BootLease;

pub use boot::BootRestore;

pub use lease::BucketLease;

pub use lease::Exclusion;

pub use lease::LeaseError;

pub use manifest::FileEntry;

pub use manifest::Manifest;

pub use manifest::SegmentEntry;

pub use manifest::MANIFEST_VERSION;

pub use restore::restore_from_s3;

pub use restore::restore_from_store;

pub use restore::RestoreS3Error;

pub use restore::RestoreS3Outcome;

pub use restore::RestoreTarget;

pub use store::build_store;

pub use store::BackupConfig;

pub use store::S3Location;

pub use worker::BackupCounters;

pub use worker::BackupHandle;

pub use worker::BackupTrigger;

pub use worker::BackupWorker;

pub use worker::WalSource;

pub use worker::SIDE_FILES;

pub use object_store;

Modules

boot

Restore on boot: a server whose data directory holds no store restores the newest restore point of a backup prefix into it before it opens the store (MUSHROOMDB_RESTORE_FROM_S3).

disk

Everything the worker reads from the store directory, as blocking calls.

lease

The bucket lease: an advisory, remote single-writer lock on a backup prefix.

manifest

The manifest: one consistent restore point, and the object keys it names.

restore

Restore a store from a backup prefix: mushroomdb restore --from s3://….

segment

Tailing wal.bin: the next byte range to ship, in whole frames.

store

Configuration from the environment, and the object store it names.

worker

The backup task: ships WAL segments, snapshots, archives and side files, and writes the manifests that name consistent restore points.

Enums

BackupError

Why a backup step or its configuration failed.

Functions

redact

text with access-key ids, signatures and other credential material replaced by [redacted]: the content of [REDACTED_ELEMENTS] XML elements, the value after each of [REDACTED_FIELDS], and anything shaped like an AWS access key id (AKIA…, ASIA…: four letters and sixteen more upper-case letters or digits).