[
https://issues.apache.org/jira/browse/PHOENIX-1058?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14093344#comment-14093344
]
James Taylor commented on PHOENIX-1058:
---------------------------------------
Wow, this is some hardcore HBase code, most of it over my head. It would be
interesting to have [~lhofhansl], [~apurtell], and [~jesse_yates] take a look
if they haven't already wrt to hindex HBase JIRA.
Given that this is isolated to local indexing and local indexing isn't really
functional without it, I'm going to pull it in and have [~mujtaba] run his
cluster-level local indexing tests over it.
As far as the Phoenix changes, one thing that caught my eye that I hope isn't
necessary is the following. This will force a scan over meta every time a local
index is used. Our approach today is to detect when the region boundaries have
changed (we can tell in our coprocessor), and then retry the chunk of work
(which may now be multiple chunks of work post-split). Any reason why this
approach wouldn't work for local indexing as well?
{code}
@Override
protected List<HRegionLocation> getAllRegions() throws SQLException {
+
context.getConnection().getQueryServices().clearTableRegionCache(tableRef.getTable().getPhysicalName().getBytes());
return
context.getConnection().getQueryServices().getAllTableRegions(tableRef.getTable().getPhysicalName().getBytes());
}
{code}
> Support index region split on it's corresponding data region split
> ------------------------------------------------------------------
>
> Key: PHOENIX-1058
> URL: https://issues.apache.org/jira/browse/PHOENIX-1058
> Project: Phoenix
> Issue Type: Sub-task
> Reporter: rajeshbabu
> Assignee: rajeshbabu
> Fix For: 5.0.0, 4.1
>
> Attachments: PHOENIX-1058.patch
>
>
> When data region split corresponding index region also should split to ensure
> co-location. At the same time we should ensure index region data split
> properly into it's daughter regions.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
