xinnyuli commented on PR #12454:
URL: https://github.com/apache/seatunnel/pull/12454#issuecomment-5992882435

   
   <blockquote>
   <p>Ran the PR's own extended IT on ba304980c71f with the same setup as 
before 
(<code>PostgresCDCIT#testPostgresCdcSnapshotOnlyAndCommittedOffsetStartupModes</code>,
 Java 11, stress-ng on, no agent, Debezium resume logging on). Run: 
https://github.com/xinnyuli/seatunnel-12382-repro/actions/runs/37292055597</p>
   <ul>
   <li>30/30 runs reached the id=15 check and passed (0 ROW_MISSING, 0 setup 
failures).</li>
   <li><strong>10/30 hit the ambiguous boundary</strong> (first LSN on restart 
== stored commit-end LSN). All 10 delivered the row: nothing was filtered as 
already processed, and the resume position equalled the stored LSN in every 
case.</li>
   </ul>
   
   run | stored lsn_proc | first LSN on restart | skipped as processed | outcome
   -- | -- | -- | -- | --
   n1 | 0/2229658 | 0/2229658 | – | PASS
   n2 | 0/222A260 | 0/222A260 | – | PASS
   n3 | 0/222A9E0 | 0/222A9E0 | – | PASS
   n6 | 0/2228860 | 0/2228860 | – | PASS
   n15 | 0/2225710 | 0/2225710 | – | PASS
   n19 | 0/222D698 | 0/222D698 | – | PASS
   n20 | 0/2224D78 | 0/2224D78 | – | PASS
   n21 | 0/2229A90 | 0/2229A90 | – | PASS
   n24 | 0/22268E0 | 0/22268E0 | – | PASS
   n28 | 0/22270D0 | 0/22270D0 | – | PASS
   
   
   <p>Since the test asserts id=15 and id=16 in the same block, PASS covers 
both. The WAL-range logging wasn't used here because the patch doesn't apply to 
the updated test (id=15 is now inserted while the job is stopped); boundary 
hits come from the Debezium resume logs instead.</p>
   <p>One caveat: this is the fixed code only. I didn't run this updated test 
against the PR base, so it doesn't by itself show the old code losing the row 
under the new test. The earlier before/after (3/30 vs 0/28 on the original 
test) covers that side.</p>
   


-- 
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