High Availability Steward is a (WIP) Go CLI and Kubernetes operator for database cluster triage, repair, backup, and restore. Pronounced like Haste·Ward or H.A. or Ha! Steward — flexible pronunciation. Backups use restic for block-level dedup, encryption, and compression.
| Engine | Database | Operator |
|---|---|---|
cnpg |
PostgreSQL | CloudNativePG |
galera |
MariaDB | mariadb-operator |
| Triage | Read-only diagnostics: pg_controldata, grastate.dat, replication status, disk usage, split-brain |
| Repair | Automated heal with pre-repair escrow backup, split-brain forensic capture, and safety gates |
| Backup / Restore | Streaming dump through restic backup --stdin — no temp files on database pods |
| Retention | Restic-style tag retention with group-aware diverged snapshot pruning |
| Operator Mode | CRD-driven scheduler watches database CRs and runs triage/repair/backup on cron |
| Bootstrap | Full Galera cluster recovery from total failure with dry-run preview |
| WAL Prune | Emergency CNPG WAL cleanup for disk-full deadlock recovery |
| Machine Output | --output json|jsonl for automation with typed envelopes, JSONL events, and --dry-run support |
| Topic | |
|---|---|
| CLI Reference | Subcommands and global flags |
| Examples | CLI usage examples for every subcommand |
| Backups | Backup model, snapshot tags, retention |
| Safety Gates | Repair and bootstrap safety matrices |
| Operator Mode | CRDs, annotations, scheduler |
| Architecture | Engine repair flows, backup streaming |
| File | |
|---|---|
| Kubernetes Job | Ad-hoc Job manifest for CLI operations |
Distributed under the AGPL-3.0-only License. See LICENSING.md for commercial licensing.