shuke987 opened a new pull request, #69013:
URL: https://github.com/apache/doris/pull/69013

   ### What problem does this PR solve?
   
   Issue Number: None
   
   Related PR: #69008 (independent gzip upload improvement)
   
   Problem Summary:
   
   Litefuse verification fetches trace detail and then unconditionally 
downloads the same observations through the paginated API. When detail already 
contains all observation IO, these extra reads consume the shared 120-second 
verification budget and can leave later traces unpolled.
   
   Extract the existing completeness checks into a shared validator. Validate 
trace-detail observations first and finish that trace immediately when all 
checks pass. Incomplete detail still falls back to the existing legacy/v2 
observation readers. The observation count, IDs, duplicates, IO, subagent 
expected IDs, root and agent-message context checks are unchanged.
   
   Verification remains enabled with the same shared timeout and polling 
behavior. The workflow's existing `continue-on-error: true` still isolates 
telemetry failures from review status. This PR changes only the exporter and 
its tests; it is independent of gzip and introduces no concurrency.
   
   Validation:
   - Reproduced on the original implementation: a deterministic nine-trace test 
with eight seconds of simulated latency per request exhausts the 120-second 
budget and leaves the last trace unpolled. The fix verifies all nine traces 
with nine detail requests in 72 simulated seconds.
   - 42 exporter tests pass, including incomplete detail fallback, 
IDs/duplicates/IO, agent-message context, existing pagination fallbacks and 
deadline checks. The regression tests fail on the original implementation.
   - An AST comparison confirms the extracted validation body is unchanged.
   - Offline replay of five saved traces (1,850 observations) from run 
38323941693 verifies the same traces while reducing GET requests from 26 to 5. 
Serialized replay response bytes fall from 40,094,490 to 20,054,297. These are 
replay measurements, not a live latency benchmark.
   - Live read-only check of that run's 315-observation main trace passes 
complete verification with one detail GET (2.319 seconds for this sample).
   - The two affected files merge cleanly with #69008; all 44 combined exporter 
tests pass. No whitespace errors.
   
   ### Release note
   
   None
   
   ### Check List (For Author)
   
   - Test
       - [ ] Regression test
       - [x] Unit Test
       - [x] Manual test (historical trace replay and live read-only 
verification)
   
   - Behavior changed:
       - [ ] No.
       - [x] Yes. Complete trace detail skips redundant observation downloads.
   
   - Does this need documentation?
       - [x] No.
       - [ ] Yes.
   
   ### Check List (For Reviewer who merge this PR)
   
   - [ ] Confirm the release note
   - [ ] Confirm test cases
   - [ ] Confirm document
   - [ ] Add branch pick label
   


-- 
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