ShummGen commented on issue #63609:
URL: https://github.com/apache/doris/issues/63609#issuecomment-4533265380

   Adding more details about the affected tablets.
   
   The corrupted tablets confirmed from BE-side logs are:
   
   - `1775727969562`
   - `1775727969030`
   - `1775728203127`
   
   Known facts for these tablets:
   - they were all created in the same narrow time window around `May 14 09:40`
   - cumulative compaction repeatedly failed on these tablets
   - the corruption symptoms are `checksum mismatch` and `ZSTD decompression 
failed`
   
   Related BE log examples:
   
   ```text
   W20250525 09:10:15.234567 86955 cumulative_compaction.cpp:890] failed to do 
cumulative compaction. tablet=1775727969562, err=Corruption: checksum mismatch
   W20250525 09:10:15.234890 86955 cumulative_compaction.cpp:890] failed to do 
cumulative compaction. tablet=1775727969030, err=Corruption: checksum mismatch
   W20250525 09:10:15.235123 86955 cumulative_compaction.cpp:890] failed to do 
cumulative compaction. tablet=1775728203127, err=Corruption: checksum mismatch
   One example corrupted segment file is:
   
   
/home/doris_test_local/be-storage/data/504/1775727969562/29255074/02000000000ac7103d41342af722a463b0a66dca056a21a2_2.dat
   For that file, we observed:
   
   actual checksum = 3512177969
   expected checksum = 2503684114
   We also observed another BE node with 36 corrupted tablets created during 
the same May 13-14 time window, which suggests this was a historical 
cluster-wide corruption event rather than a single-tablet isolated case.
   
   The main issue we want to report here is that when Doris later reads these 
corrupted tablets, BE crashes with SIGSEGV instead of failing gracefully.


-- 
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