[jira] Updated: (HADOOP-2377) [hbase] Holding open MapFile.Readers is expensive, so use less of them

["stack (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/HADOOP-2377?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

stack updated HADOOP-2377:
--------------------------

    Attachment: bigger.patch

Made regions 4 times larger by default.  They're now 256M instead of 64M.  Also upped memcache size threshold from 16M to 64M.  Tests show splits taking same time -- split time shouldn't be effected by time -- and compactions taking about a third longer: 20-30 seconds instead of 8-10 seconds (They run in background).   Running the PerformanceEvaluation test, before we ended up w/ 19 regions doing write test-- now 5.

M  src/contrib/hbase/conf/hbase-default.xml
    Upped flush.size and max.filesize
M  src/contrib/hbase/src/test/org/apache/hadoop/hbase/PerformanceEvaluation.java
    Don't delete table on end of test and don't do auto-upload of content....
    just fail if none there.
M  src/contrib/hbase/src/java/org/apache/hadoop/hbase/HStoreFile.java
    readSplitInfo: Looks like a leak.
M  src/contrib/hbase/src/java/org/apache/hadoop/hbase/HConstants.java
    Up region MAX_FILE_SIZE to 256M.
M  src/contrib/hbase/src/java/org/apache/hadoop/hbase/HRegion.java
    Up flush size to 64M



> [hbase] Holding open MapFile.Readers is expensive, so use less of them
> ----------------------------------------------------------------------
>
>                 Key: HADOOP-2377
>                 URL: https://issues.apache.org/jira/browse/HADOOP-2377
>             Project: Hadoop
>          Issue Type: Improvement
>          Components: contrib/hbase
>            Reporter: stack
>             Fix For: 0.16.0
>
>         Attachments: bigger.patch
>
>


-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.