[ https://issues.apache.org/jira/browse/HBASE-17057?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15703800#comment-15703800 ]
Ashu Pachauri edited comment on HBASE-17057 at 11/29/16 1:59 AM: ----------------------------------------------------------------- I had a first look at the code, things never are straightforward in the land of compactions. 1. Whether we can drop cache behind reads/writes for compactions is tied to whether we should throttle compactions. Not sure what dropping page cache has to do with throttling compactions. [~eclark] , can you shed a little light on why this is the case? (Since you added this piece of code in HBASE-14098 ) 2. It seems that we just ignore drop cache hint for all compactions on master branch. I'll dig more and try to find why the behavior was modified on master. was (Author: ashu210890): I had a first look at the code, things never are straightforward in the land of compactions. 1. Whether we can drop cache behind reads/writes for compaction is tried to whether we should throttle compactions. Not sure what dropping page cache has to do with throttling compactions. [~eclark] , can you shed a little light on why this is the case? (Since you added this piece of code in HBASE-14098 ) 2. It seems that we just ignore drop cache hint for all compactions on master branch. I'll dig more and try to find why the behavior was modified on master. > Minor compactions should also drop page cache behind reads > ---------------------------------------------------------- > > Key: HBASE-17057 > URL: https://issues.apache.org/jira/browse/HBASE-17057 > Project: HBase > Issue Type: Improvement > Components: Compaction > Reporter: Ashu Pachauri > Assignee: Ashu Pachauri > > Long compactions currently drop cache behind reads so that they don't pollute > the page cache but short compactions don't do that. The bulk of the data is > actually read during minor compactions instead of major compactions, and > thrashes the page cache since it's mostly not needed. > We should drop page cache behind reads for minor compactions too. -- This message was sent by Atlassian JIRA (v6.3.4#6332)