[
https://issues.apache.org/jira/browse/PHOENIX-4988?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16661761#comment-16661761
]
Lars Hofhansl commented on PHOENIX-4988:
----------------------------------------
Do we need to do the RAW scan here?
Much of the work we're doing (including keeping the right delete family marker)
is already handled in HBase if we let it (i.e. not do a RAW scan).
I will point out that there might be another snag: The second family marker we
see might actually be older than the first, in which case we need to keep the
second one. (That can happen when a family delete marker is placed later - such
that it ends up in a different HFile, but with an earlier timestamp. HBase
handles that as well - see ScanDeleteTracker.) Not something we need to worry
about here, I think, since only Phoenix places these delete marker and it won't
backdate deletes.
> Incorrect index rowkey generated when updating only non-indexed columns after
> a delete
> --------------------------------------------------------------------------------------
>
> Key: PHOENIX-4988
> URL: https://issues.apache.org/jira/browse/PHOENIX-4988
> Project: Phoenix
> Issue Type: Bug
> Affects Versions: 4.14.0
> Reporter: Vincent Poon
> Assignee: Vincent Poon
> Priority: Critical
> Fix For: 4.14.1
>
> Attachments: PHOENIX-4988.v1.4.x-HBase-1.4.patch,
> PHOENIX-4988.v1.master.patch
>
>
> The following steps result in an incorrect index rowkey being generated after
> an index update to a non-indexed column.
> create table test (k VARCHAR NOT NULL PRIMARY KEY, v1 VARCHAR, v2 VARCHAR)
> create index test_ind ON test (v2)
> upsert into test (k,v1,v2) VALUES ('testKey','v1_1','v2_1');
> delete from test;
> upsert into test (k,v1,v2) VALUES ('testKey','v1_2','v2_2');
> delete from test;
> upsert into test (k,v1) VALUES ('testKey','v1_3');
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
