rangareddy commented on issue #17355:
URL: https://github.com/apache/hudi/issues/17355#issuecomment-5351238030
This issue was reviewed as part of the JIRA-migrated backlog triage
(HUDI-8825).
**Findings: needs the attached reproduction run against a current build
before it can be dispositioned.**
This report is unusually complete - a full SQL suite with a MOR table,
`GLOBAL_SIMPLE` index, `hoodie.index.global.index.enable = 'true'`, a `DOUBLE`
salary column and cross-partition updates through `MERGE INTO` - and it states
the impact precisely ("MIT does not work when column type is double in some
cases. No data corruption."). That is more than enough to act on; what it needs
is a current-version verdict.
I did not want to call this one from code. The `MERGE INTO` and global-index
paths have been substantially rewritten since the 0.14.1 line you observed it
on - table versions 8 and 9, the `BufferedRecordMerger` family replacing the
old payload merge path, and the metadata/indexer abstraction introduced in
#18348 and #18372. With that much churn, neither "still broken" nor "silently
fixed" is assertable by reading, and asserting either would be a guess.
The good news is the repro is self-contained: the only environment-specific
piece is the `s3a://` location, which can be swapped for a local path. Running
the suite against 1.2.x would settle this immediately.
Worth capturing in the ticket when someone does: whether the failure is
specific to `DOUBLE`, to the global index, or to the cross-partition update,
since the description suggests all three are needed to trigger it.
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]
