ranade-oss makes staging rehearsals fail closed
This fork is turning deployment debugging into a more controlled rehearsal, with stronger guardrails before anything can reach production-adjacent systems.
The change makes staging checks run against the same rehearsal setup expected by the web product, reducing the risk that a green deployment test proves the wrong thing. Runs are also queued rather than overlapping, and receive predictable temporary names so an interrupted test can be found and recovered.
- Read-only checks verify database and storage access before a rehearsal can proceed.
- Credential checks distinguish newer opaque database keys from older token-based credentials, applying the right access rules to each.
- Deployment probes confirm the intended web image and login check are actually live.
- Recovery, cleanup and evidence upload are reported separately, with success withheld until cleanup is complete.
The real staging run still needs an operator to trigger and review it manually.
So what Legal-tech teams with sensitive production data should care because this makes release rehearsal more auditable and less likely to turn a debugging exercise into an operational incident.
Spotted something wrong? Or know the PR text has fresher detail than the writeup above?