xiangfu0 opened a new pull request, #19193:
URL: https://github.com/apache/pinot/pull/19193

   ## Summary
   - run segment-commit and consuming-transition recovery scenarios in one 
shared fixture
   - preserve the standard one-server topology and the dedup 
two-server/two-replica topology
   - share one Kafka publication and fixed reference table per topology
   - isolate each fault with a unique table/schema and table-keyed immutable 
server config
   - force segment-level validation on every explicit recovery and delay the 
scheduled validator
   - remove the two redundant consuming-transition subclasses
   
   ## Why
   The four previous classes paid four complete ZooKeeper, Kafka, controller, 
broker, and server lifecycles. Their behavior only differs by fault mode and 
dedup topology.
   
   In successful reference workflow run 31227570484, the four class times 
totaled 305.6s. The consolidated local run reports 122.6s across the two 
surviving classes, a directional 60% reduction. The draft CI run will provide 
the apples-to-apples hosted measurement.
   
   This change leaves the existing two integration-test jobs unchanged and is 
independent of #19190 and #19192.
   
   ## Correctness safeguards
   - non-dedup still requires exact ERROR counts of 10 and 2
   - dedup still uses two servers, two replicas, and requires injected errors
   - both scenarios retain the nonempty ERROR, no ERROR after recovery, no 
OFFLINE after recovery, and six-field DONE metadata comparisons
   - scenario cleanup waits for ExternalView and table-data-manager removal and 
preserves primary failures
   
   ## Testing
   - PauselessRealtimeIngestionSegmentCommitFailureTest: 2 tests, 0 
failures/errors/skips, 76.38s
   - PauselessDedupRealtimeIngestionSegmentCommitFailureTest: 2 tests, 0 
failures/errors/skips, 46.18s
   - pinot-integration-tests test-compile
   - Spotless
   - Checkstyle
   - license format/check
   - git diff --check
   - independent correctness review: no findings


-- 
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]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to