Resources

Recovery runbook

Scope, precautions and acceptance checks. Recovery time remains unmeasured until an isolated drill.

Recovery scope and evidence

Provider daily backups cover database schema and data, not storage files. Daily snapshots are retained for roughly 14 days. Point-in-time recovery is not available; use the nearest earlier daily snapshot.

The actual snapshot inventory has not been verified here. Before relying on recoverability, the owner must record the newest UTC snapshot timestamp and carry out an isolated restore drill. Company JSON exports are selected application data, not full database or account backups.

RPO and RTO

Planning RPO: up to 24 hours between daily snapshots, subject to the actual snapshot evidence. This is not a verified recovery guarantee. A more recent company export can reduce loss only for the records it includes.

RTO is unmeasured. Allow for snapshot restoration, sign-in/provider configuration, file recovery and application checks. A database restore may make the database unavailable for a few minutes, but total recovery time must be measured during a drill; no SLA is claimed.

Before restoration

Name an incident owner, stop affected writes and record the selected backup timestamp, current application version and affected company. Export the current database state first and download storage files separately.

A snapshot restore is permanent: records written after the snapshot are lost, and the schema reverts with the data. Restore into a separate environment first when testing recovery; do not use a production restore as a drill.

Company-record recovery

For a supported deleted record, review its Recycle Bin snapshot, expiry and related records, then restore with administrator authorization. Check parent and child records and record any failures.

For a company JSON restore, verify the source company, table manifest and application schema, take a safety export, choose skip-existing or update deliberately, and inspect every result. Do not import untrusted documents or cross-company identifiers. Stop on incomplete safety export or any partial restore; investigate before retrying.

Full recovery and acceptance

The owner exports the whole database using the hosting account’s project-data export and downloads storage separately. The export includes user accounts/password hashes but excludes files, code, secrets and sign-in-provider configuration. Reconfigure those at a separate destination before a recovery drill.

After restoration, test email/password sign-in and authenticator access, company isolation, employees and child records, attendance/leave totals, payroll, audit records and file downloads. Compare counts against the recorded manifest. Fix schema mismatches with a separately approved migration, not ad-hoc table changes.

  • Record measured RPO/RTO, snapshot/export age and missing files.
  • Obtain owner acceptance before resuming writes.
  • Preserve incident evidence; never describe a restore as reversible.