Plan-based Postgres backup baseline
Resistro Cloud backs up Postgres databases on a plan-based cadence for small self-managed Postgres setups in the defined ICP: daily on Solo and Starter, hourly from Business.
Not covered: Postgres point-in-time recovery is proven, but not a current product promise; no automated, continuous PITR guarantee across all customer databases and no guarantee across network, host, or customer-side failure.
Scope, preconditions, evidence, customer expectation
- Scope
- Configured backup cadence for supported self-managed Postgres setups. The agent asks for backup work on a 1h default interval on every plan; the active plan decides how often a backup is accepted — daily on Solo and Starter, hourly on Business and Pro.
- Preconditions
- The customer system must be reachable, correctly attached, technically compatible, and the configured backup interval must be allowed by the active plan.
- Evidence
- Running backup jobs, heartbeat monitoring, Prometheus metrics.
- Customer expectation
- In normal operation with the default agent setup, a new backup run is created roughly each day on Solo and Starter and roughly each hour on Business and Pro.
MySQL/MariaDB: engine supports PITR, equal to PostgreSQL
MySQL and MariaDB are equal engine paths alongside PostgreSQL. The shared resistro engine supports PostgreSQL, MySQL, and MariaDB including PITR (WAL streaming / binlog replay).
Not covered: No point-in-time recovery as a current Cloud product promise, no database-version migration, and no statement for managed database platforms.
Scope, preconditions, evidence, customer expectation
- Scope
- Self-managed MySQL and MariaDB databases via the Resistro agent. The engine supports backup and PITR (binlog replay); PostgreSQL, MySQL, and MariaDB are treated equally in the engine; the Cloud path is proven by E2E for all three (backup, restore).
- Preconditions
- The customer system must be reachable, correctly attached, technically compatible, and the configured backup interval must be allowed by the active plan.
- Evidence
- The engine ships binlog archiving and MySQL/MariaDB PITR commands (documented in the open-source CLI README). Binlog archiving and PITR restore are measured in a test harness (MySQL 8.4, MariaDB 10.11: archiver over 10 minutes, restore successful). Cloud production E2E for MySQL/MariaDB is dated June 27, 2026: backup, encrypted upload, download, decrypt, restore, and checksum match.
- Customer expectation
- MySQL and MariaDB are supported on the same footing as PostgreSQL — backup, restore, and PITR (binlog replay) are engine paths and measured in the engine; backup and restore are additionally covered by the Cloud E2E.
AES-GCM encryption with customer-held key
Backup data is encrypted before storage with AES-GCM and the key stays on the customer side.
Not covered: No key recovery and no broader security guarantee beyond this storage-path claim.
Scope, preconditions, evidence, customer expectation
- Scope
- Encrypted backup storage without server-side plaintext access.
- Preconditions
- The customer must provide and manage the key correctly.
- Evidence
- Implemented encryption path; no server-side access to plaintext backup data.
- Customer expectation
- Backup data in the Resistro storage path is not stored unencrypted.
Default storage path on Hetzner Germany/EU
The default storage path uses Hetzner data centers in Germany/EU; backups are stored encrypted.
Not covered: No legal advice, no blanket privacy or risk-free claim, no promise about authority-access or third-country-transfer risks, no immunity from lawful authority requests, and no statement for customer-side third-party storage.
Scope, preconditions, evidence, customer expectation
- Scope
- Backup storage path through Hetzner infrastructure in Germany/EU.
- Preconditions
- The customer uses the standard Resistro storage path; custom storage architectures would need separate review.
- Evidence
- Current storage provider Hetzner Online GmbH; public privacy notices identify Hetzner as hosting provider.
- Customer expectation
- Storage location, encryption, AVV/TOMs, and subprocessors are documentable checks for customer use.
Daily restore test
Resistro runs a daily restore-path test as an operating check so the defined restore path does not silently break.
Not covered: No daily full emergency validation of every customer database and no application-level consistency check.
Scope, preconditions, evidence, customer expectation
- Scope
- A daily operating check of the defined restore path.
- Preconditions
- Restore-test jobs must run as scheduled and the operating environment must be available.
- Evidence
- Scheduled restore-test job, monitoring, alerting on failure.
- Customer expectation
- Resistro regularly checks that the restore mechanism still works as designed.
Stale-backup alerting after roughly one day
Resistro alerts when monitoring sees no successful backup signal for longer than the current 28h stale threshold.
Not covered: No 24/7 support, no reaction-time guarantee, and no guarantee against outages outside the Resistro monitoring path.
Scope, preconditions, evidence, customer expectation
- Scope
- Backup-failure alerting based on heartbeat and monitoring data.
- Preconditions
- Heartbeat, monitoring, and alerting infrastructure must all be functioning.
- Evidence
- Alert rules, heartbeat checks, monitoring-based alerting.
- Customer expectation
- Silent backup failure should not remain unnoticed beyond roughly one day under the monitored path.
Daily /healthz smoke test
Public product paths are checked daily for basic reachability.
Not covered: No uptime guarantee and no comprehensive end-to-end guarantee for all customer functions.
Scope, preconditions, evidence, customer expectation
- Scope
- A daily smoke test against the public health endpoint.
- Preconditions
- The infrastructure and health endpoint must be reachable.
- Evidence
- Daily smoke test against `/healthz`.
- Customer expectation
- Basic reachability of central public product paths is checked every day.
Plan-based RPO
The target RPO in normal operation is around one day on Solo and Starter and around one hour on Business and Pro.
Not covered: No hard uptime or data-loss guarantee for failures between backup windows or during extended disruption.
Scope, preconditions, evidence, customer expectation
- Scope
- A target derived from the plan-based backup cadence and monitoring.
- Preconditions
- Backups need to run successfully without a longer disruption chain.
- Evidence
- Plan-based backup cadence plus monitoring, without hard guarantee language.
- Customer expectation
- Potential data loss in normal operation is bounded to roughly one day on Solo and Starter and roughly one hour on Business and Pro.
Manual operator-assisted restore
Restore is a controlled manual operation with operator support, not a self-service workflow.
Not covered: No support guarantee, no fixed RTO, and no self-service restore portal.
Scope, preconditions, evidence, customer expectation
- Scope
- Manual restore handling when a valid backup exists and operational restore is possible.
- Preconditions
- A valid backup must exist and the restore can be carried out operationally.
- Evidence
- Restore process and daily restore-path test.
- Customer expectation
- Restore is a supervised operational process rather than a push-button action.
Customer-triggered restore drills
Customers can trigger restore drills on demand to verify the restore path for their own databases.
Not covered: A drill is not a production restore and does not constitute a backup guarantee beyond the verified restore path.
Scope, preconditions, evidence, customer expectation
- Scope
- On-demand drill triggering against the customer's own backup data.
- Preconditions
- A valid backup must exist for the selected database.
- Evidence
- Drill trigger endpoint, drill result logging, restore-evidence PDF generation.
- Customer expectation
- A restore drill verifies that the restore path works for the customer's data at the time of the drill.
Restore-evidence PDF
After a restore drill, Resistro generates a downloadable restore-evidence PDF documenting the result.
Not covered: Not a legal certification, not a compliance certificate, not production-restore evidence, and not a PITR promise.
Scope, preconditions, evidence, customer expectation
- Scope
- PDF export of drill result as audit evidence.
- Preconditions
- A restore drill must have been completed successfully.
- Evidence
- PDF generated from drill result data including timestamp, database, and outcome.
- Customer expectation
- The PDF provides structured evidence of a completed restore drill for internal review or customer evidence.
Operator context
Who runs this
Alexander Renz is a Linux administrator, the founder of Resistro Cloud, and a small business owner since April 28, 2026. He operates the infrastructure himself and keeps the public promise intentionally bounded to what is real in production today.