DanielLeens commented on issue #12019:
URL: https://github.com/apache/seatunnel/issues/12019#issuecomment-5478145627

   Thanks for the report. I rechecked the current Oracle CDC split path on 
`dev` before replying.
   
   The stack trace here is important: the failure is happening while Oracle CDC 
is generating snapshot splits, and on the current source the concrete statement 
that dies in this path is `jdbc.execute(analyzeTable)` inside 
`OracleUtils.queryApproximateRowCnt(...)`, not the downstream sink path or the 
LogMiner streaming phase.
   
   Today the relevant defaults are:
   
   1. `use_select_count = false`;
   2. `skip_analyze = false`;
   3. so the default path runs `analyze table ... compute statistics for table` 
before it reads `NUM_ROWS`.
   
   Because of that, I would not call this a confirmed SeaTunnel product bug 
yet. First we need to separate "the analyze statement is timing out / losing 
its Oracle session" from "any row-count path fails".
   
   Could you please retry with these two controlled variants and report exactly 
which one fails?
   
   1. `skip_analyze = true` while keeping `use_select_count = false` so the 
split path skips `analyze table` and reads `NUM_ROWS` directly.
   2. `use_select_count = true` so the split path skips `analyze table` 
entirely and uses `select count(*) from ...`.
   
   If variant 1 or 2 succeeds, the failure is likely specific to the 
analyze/statistics path or its runtime duration. If both still fail with the 
same `No more data to read from socket`, then we should treat it as a broader 
Oracle JDBC/session-stability issue in split enumeration.
   
   When you reply, please also include:
   
   1. the exact Oracle database version;
   2. whether the failure still reproduces on the same table with `parallelism 
= 1`;
   3. any server-side SQL/socket timeout settings that may close a long-running 
Oracle session.
   
   That will let maintainers narrow this much faster without guessing.
   


-- 
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