[
https://issues.apache.org/jira/browse/HBASE-10884?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13965674#comment-13965674
]
Lars Hofhansl commented on HBASE-10884:
---------------------------------------
Maybe in 0.94 we should invert the default (leave it at false and allow
enabling this)...?
I'm fine either way. In 0.94 we should limit the element of the surprise. But
which of the following two is more suprising?
# scanning via REST and the Java API use different defaults
# in 0.94.18 scanning via REST did not cache, but in 0.94.19 it does by default.
I guess I'm fine either way.
Any opinions?
> [REST] Do not disable block caching when scanning
> -------------------------------------------------
>
> Key: HBASE-10884
> URL: https://issues.apache.org/jira/browse/HBASE-10884
> Project: HBase
> Issue Type: Improvement
> Affects Versions: 0.98.1, 0.96.1.1, 0.94.18
> Reporter: Andrew Purtell
> Assignee: Andrew Purtell
> Attachments: HBASE-10884.patch
>
>
> The REST gateway pessimistically disables block caching when issuing Scans to
> the cluster, using Scan#setCacheBlocks(false) in ScannerResultGenerator. It
> does not do this when issuing Gets on behalf of HTTP clients in
> RowResultGenerator. This is an old idea now, the reasons for doing so lost
> sometime back in the era when HBase walked the earth with dinosaurs (< 0.20).
> We probably should not be penalizing REST scans in this way.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
