hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Andrew Purtell (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HBASE-1562) How to handle the setting of 32 bit versus 64 bit machines
Date Mon, 22 Jun 2009 17:34:07 GMT

    [ https://issues.apache.org/jira/browse/HBASE-1562?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12722699#action_12722699
] 

Andrew Purtell commented on HBASE-1562:
---------------------------------------

I am seeing HeapSize/ClassSize related failures on a 64 bit platform also, in TestHFile:

{noformat}
2009-06-22 17:31:59,860 DEBUG [main] util.ClassSize(147): Primitives 8, arrays 3
, references(inlcuding 16, for object overhead) 3, refSize 4, size 72
------------- ---------------- ---------------

Testcase: testTFileFeatures took 0.359 sec
Testcase: testMetaBlocks took 0.088 sec
Testcase: testNullMetaBlocks took 0.05 sec
Testcase: testCompressionOrdinance took 0.027 sec
Testcase: testComparator took 0.046 sec
Testcase: testHeapSizeForBlockIndex took 0.245 sec
        FAILED
expected:<72> but was:<128>
junit.framework.AssertionFailedError: expected:<72> but was:<128>
        at org.apache.hadoop.hbase.io.hfile.TestHFile.testHeapSizeForBlockIndex(TestHFile.java:273)
{noformat}


> How to handle the setting of 32 bit versus 64 bit machines
> ----------------------------------------------------------
>
>                 Key: HBASE-1562
>                 URL: https://issues.apache.org/jira/browse/HBASE-1562
>             Project: Hadoop HBase
>          Issue Type: Bug
>    Affects Versions: 0.20.0
>            Reporter: Erik Holstad
>            Assignee: Erik Holstad
>             Fix For: 0.20.0
>
>
> After adding the tests to verify the correctness of the HeapSize calculations the question
of where to set the type of machines that are in the cluster arose.
> I would think that most people are using 64 bit machines but we still need to support
the use of 32 bit. So the way I see it we can solve this problem in two ways,
> we can either have a settable parameter the the user sets when starting up the cluster
or we can try to figure it out ourselves. I think that the second solution would
> be the best, to make it as easy as possible on the user. 
> That means that we need to add extra sizes to HeapSize and maybe even to Bytes.
> Thoughts, comments?

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


Mime
View raw message