[
https://issues.apache.org/jira/browse/PHOENIX-3821?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15991819#comment-15991819
]
James Taylor commented on PHOENIX-3821:
---------------------------------------
Besides adding a new config in QueryServices and a default value in
QueryServicesOptions, this change will be isolated to UpsertCompiler, in this
block:
{code}
runOnServer = isAutoCommit && !table.isTransactional()
&& !(table.isImmutableRows() &&
!table.getIndexes().isEmpty())
&& !select.isJoin() &&
table.getRowTimestampColPos() == -1;
{code}
Prior to PHOENIX-3271, we also used the {{sameTable}} boolean, set a few lines
above, in this check. If we conditionally check this based on our new config,
that should be enough to force the statement to run on the client.
> Provide config to disable distributing upsert/select across cluster
> -------------------------------------------------------------------
>
> Key: PHOENIX-3821
> URL: https://issues.apache.org/jira/browse/PHOENIX-3821
> Project: Phoenix
> Issue Type: Bug
> Reporter: James Taylor
> Assignee: Alex Araujo
>
> We should provide a way to configure the system such that an UPSERT/SELECT is
> run the way it was before PHOENIX-3271 (i.e. not distributed across the
> cluster if the source and target tables are different).
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
