vyommani opened a new pull request, #1208:
URL: https://github.com/apache/ranger/pull/1208

   ## What changes were proposed in this pull request?
   
   Strengthens how Ranger Admin protects certain encrypted configuration values 
(e.g. service
   connection passwords) at rest, and migrates existing data to the improved 
approach. New writes
   get the improved protection immediately; a one-time patch upgrades data 
written before this
   change.
   
   ## How was this patch tested?
   
   
   Ran end-to-end against a real running instance 
(`dev-support/ranger-docker`), not just unit
   tests, using genuine legacy-format data:
   
   - **Migration**: all genuinely-legacy rows converted; each row's stored 
value differs from its
     pre-migration value (confirms a real decrypt/re-encrypt, not a relabel); 
zero failures.
   - **HA mismatch**: pointed one node at a different key than the one data was 
migrated under —
     fails closed with an actionable error instead of a silent bad-password 
auth failure; reverting
     the key restores correct decryption.
   - **Idempotency**: re-running the migration patch against already-migrated 
data is a no-op, no
     failures.
   - Full `mvn test` / project build verification is still pending — this pass 
was done in an
     environment without Maven Central access, so verification so far is 
against the modified
     classes compiled and run directly, not through the project's own build.
   
   ### Open questions for reviewers (flagged in-code, not silently decided)
   
   1. Migration patch runs single-pass/single-transaction (matches existing 
precedent in this
      codebase) — acceptable at very large scale, or should it batch?
   2. The new consistency checker WARNs and lets startup continue on a key 
mismatch rather than
      failing it outright, to avoid turning a legitimate transient state (e.g. 
an in-progress key
      rotation) into an outage. Right trade-off, or should a hard mismatch 
block startup?
   3. The new read-path branch relies on the new decrypt call throwing reliably 
on a wrong key,
      rather than repeating the existing byte-exact re-encrypt-and-compare 
self-check the legacy
      branches use. Is that reliance justified as-is?
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to