rangareddy commented on issue #17136:
URL: https://github.com/apache/hudi/issues/17136#issuecomment-5366100521

   This issue was reviewed as part of the JIRA-migrated backlog triage 
(HUDI-9686).
   
   **Findings: needs a reproduction against a current build before it can be 
dispositioned.**
   
   The report is concrete: a COW table created on 0.13.1 with 50 commits (so 
part of the timeline is archived), auto-upgraded to 0.14.1, then failing with a 
NullPointerException on the first write after moving to 1.0.2. Reproducing it 
looks tractable - the reporter supplied a full gist reproduction 
(https://gist.github.com/deepakpanda93/f33c2a3a7b9a200f93a3be351ba63b86), and 
the archived-timeline detail is the important part - it is why 50 commits are 
specified rather than a handful.
   
   I have not called this from the source. The surrounding code has been 
rewritten substantially since this was filed - table versions 8 and 9, the 
`BufferedRecordMerger` family replacing payload-based merging, the 
file-group-reader migration, the indexer abstraction in #18348 / #18372, and 
the `hudi-common` package moves in #19193 / #19195. Inferring either "still 
broken" or "silently fixed" through that much churn would be a guess.
   
   There is also no commit anywhere in the tree referencing HUDI-9686, so if it 
was fixed it was incidental rather than deliberate - another reason to verify 
rather than assume.
   
   Multi-version upgrade paths are exactly where regressions hide, and this one 
crosses two major boundaries. Worth capturing which component NPEs: an 
archived-timeline read on the 0.13.1-era layout is the most likely culprit, 
since table version 5-era archived blocks are parsed by newer readers.
   
   Keeping this open.
   


-- 
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