ann created HUDI-7030:
-------------------------
Summary: Log reader data lost as that not consistent behavior in
timeline's containsInstant
Key: HUDI-7030
URL: https://issues.apache.org/jira/browse/HUDI-7030
Project: Apache Hudi
Issue Type: Bug
Affects Versions: 0.14.0
Reporter: ann
Attachments: image-2023-11-03-19-48-29-441.png,
image-2023-11-03-19-49-22-894.png, image-2023-11-03-19-50-11-849.png,
image-2023-11-03-19-58-39-495.png, image-2023-11-03-20-06-00-579.png,
image-2023-11-03-20-06-13-905.png, image-2023-11-03-20-07-30-201.png
Log reader is filtered all log data blocks which come from inflight instant.
!image-2023-11-03-19-49-22-894.png!
*containsInstant* return false when input instant's timestamp is not equal as
anyone instant timestamp in inflight timeline.
!image-2023-11-03-20-07-30-201.png!
But now, in timeline's *containsInstant* that input is instant's timestamp, it
would return true.
!https://issues.apache.org/jira/secure/temporaryattachment/e53a9a4d763501eff6baeff2832c2a2c44a30713/temp2778457713180842709_image-2023-11-03-20-06-13-905.png!
When input is the instant with default_millis_ext, instant's timestamp is less
than someone instant timestamp in timeline.
!image-2023-11-03-19-50-11-849.png!
In finally, log reader skipped the completed delta commit instant and caused
data lost.
!image-2023-11-03-19-58-39-495.png!
I think timeline's containsInstant should have consistent behavior and update
containsOrBeforeTimelineStarts to containsInstant
!image-2023-11-03-19-48-29-441.png!
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
