[
https://issues.apache.org/jira/browse/PHOENIX-1100?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14067752#comment-14067752
]
Jan Fernando commented on PHOENIX-1100:
---------------------------------------
Awesome [~jamestaylor]! Thank you. I'll apply the patch locally and try it out
with Query More.
> Upsert Select with Sequence and Order By doesn't result in sequence getting
> assigned based on requested ordering
> -----------------------------------------------------------------------------------------------------------------
>
> Key: PHOENIX-1100
> URL: https://issues.apache.org/jira/browse/PHOENIX-1100
> Project: Phoenix
> Issue Type: Bug
> Affects Versions: 3.0.0, 4.0.0, 5.0.0
> Reporter: Jan Fernando
> Assignee: James Taylor
> Fix For: 5.0.0, 3.1, 4.1
>
> Attachments: PHOENIX-1100.patch, PHOENIX-1100_v2.patch
>
>
> I working with Query More implementation I noticed that once we had
> sufficient parallelization in the SELECT portion of UPSERT...SELECT due to
> Salting or region splits the results were not returned in the correct order.
> The root cause appeared to be due to the fact that we are using sequences to
> build a cursor of data for our Query More implementation. What appears to be
> happening is that as we get the next sequence value from the SequenceManager
> during result processing parallel threads all increment the sequence and
> therefore the sequence values reflect the processing order of each iterator
> versus the actual order specified by the order by. The expectation is that
> the sequence value should reflect the requested order specified by the ORDER
> BY versus the processing order.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
