Skip to content
Back to products
Controlled pilot

InfraShift

Controlled database migration and verified full backups

Deterministic, self-hosted database migration and verified full-backup tooling across PostgreSQL, MySQL, SQL Server, and MongoDB.

Discuss a project

What it is

InfraShift is a self-hosted Rust command-line tool for deterministic database migration and verified full backups. It runs where your data runs, with no SaaS control plane and no external orchestration.

Database migrations and full restores you can rehearse: derive a plan, execute it with bounded concurrency and resumable checkpoints, validate the result, and keep cutover under operator control.

Capabilities

Analyse and plan

Analyse source and target, derive a dependency-aware plan, normalise schemas, and select pair-specific type mappings with fail-closed behaviour for unsupported conversions.

Execute and resume

Copy schema and data with bounded concurrency, retries, and atomic local checkpoints. Interrupted jobs resume from local state.

Validate

Validate row counts, checksums where supported, constraints, and PostgreSQL sequences.

Change data capture

For MySQL to MySQL only: interactive CDC with an explicit barrier, DDL freeze, drain, validation, and operator-controlled cutover.

Backup and restore

Create vendor-native full backups, chunk and optionally encrypt them, inspect, verify, list, and prune them, and restore to a new or empty same-engine target.

Who it is for

  • Database administrators.
  • Platform and SRE teams.
  • Migration consultants and delivery teams.
  • Organisations that require controlled or restricted-network execution.

These are supported personas and designed audiences, not evidence of current customers.

Technical proof

  • All 16 directed bulk-migration combinations across the four supported engines.
  • Directional type mapping with fail-closed behaviour for unsupported conversions.
  • No Kafka and no external control plane required.
  • Local resumable state and bounded retries.
  • Backup targets: local and NFS, S3-compatible storage, Azure Blob, and Google Cloud Storage.
  • AES-GCM authenticated encryption with integrity verification.
  • Structured, credential-redacted logs and an optional Prometheus metrics endpoint.
  • Separate Rust crates for CLI, migration orchestration, canonical model, connectors, validation, CDC, backup, and metrics.

Availability and scope

Controlled pilot

InfraShift is explicitly pre-1.0. A dated qualification for commit 7bdbac9 on 14 August 2026 exercised all four backup engines, all 16 bulk routes, MySQL CDC, interrupted-job resume, an approximately 10 GiB and 52 million-row PostgreSQL backup and restore, and negative cases. That report concludes GO for pull request and NO-GO for an uncontrolled pilot until an exact-head rerun, and it rejects using InfraShift as a replacement for an established point-in-time-recovery system. Results apply to that commit, date, and environment only.

Current limitations

  • Migration is not zero downtime in the general case.
  • Change data capture is MySQL to MySQL only. There is no all-engine or cross-engine CDC.
  • Not general availability, not production-ready, not enterprise-ready.
  • No REST API, SaaS control plane, or distributed orchestration.
  • No incremental backup, point-in-time recovery, log shipping, or arbitrary in-place restore. Restore targets a new or empty same-engine database.
  • No universal recovery-time, throughput, lag, or cutover figures.
  • Not a single static binary: native database tools and libraries are dependencies.
  • The optional metrics endpoint has no application authentication and must remain on a trusted network.
  • No cloud-provider or compliance certification.