In any case, this thread was about 0.94 and JDK7, which should work fine (we're 
using a fully secured environment - in fact there is a bug with kerberos cross 
realm authentication with JDK*6*)

Or are you saying the Hadoop2.2 with JDK6 passes all tests and Hadoop2.2 with 
JDK7 doesn't?


The tests do not all pass with Hadoop 2 in 0.94, that is known, maybe we should 
fix that.
At Salesforce we use HBase with Hadoop 2.0.x (HDFS and Yarn) and I can confirm 
that it all works.

-- Lars



________________________________
 From: Ted Yu <[email protected]>
To: "[email protected]" <[email protected]>; lars hofhansl 
<[email protected]> 
Sent: Saturday, January 11, 2014 7:19 PM
Subject: Re: 0.94 and JDK7
 

According to the link from my previous email, 0.94.x can run on hadoop 2.2.0

I regenerated HBaseProtos.java and ErrorHandlingProtos.java using protoc
2.5.0
TestForeignExceptionSerialization passed after that.

Cheers



On Sat, Jan 11, 2014 at 6:12 PM, lars hofhansl <[email protected]> wrote:

> Which version of Hadoop 2?
> Hadoop 2.2.x is known not to work, due to mismatch of protobuf (Hadoop
> 2.2.0 upgrades to 1.5)
>
> -- Lars
>
>
>
> ________________________________
>  From: Ted Yu <[email protected]>
> To: "[email protected]" <[email protected]>
> Sent: Saturday, January 11, 2014 5:34 PM
> Subject: Re: 0.94 and JDK7
>
>
> I followed http://hbase.apache.org/book.html#d2907e640 to compile 0.94
> against hadoop-2 using jdk7
>
> I got the following errors running test suite :
>
>
>
> testFailingCreateTable(org.apache.hadoop.hbase.rest.client.TestRemoteAdminRetries):
> This is supposed to be overridden by subclasses.
>
>
> testSplitSampler(org.apache.hadoop.hbase.mapreduce.hadoopbackport.TestInputSampler):
> Failed getting constructor
>
>
> testIntervalSampler(org.apache.hadoop.hbase.mapreduce.hadoopbackport.TestInputSampler):
> Failed getting constructor
>
>
> testGetScanner(org.apache.hadoop.hbase.rest.client.TestRemoteHTableRetries):
> This is supposed to be overridden by subclasses.
>
>
> testSingleRowPut(org.apache.hadoop.hbase.rest.client.TestRemoteHTableRetries):
> This is supposed to be overridden by subclasses.
>
>
> testMultiRowPut(org.apache.hadoop.hbase.rest.client.TestRemoteHTableRetries):
> This is supposed to be overridden by subclasses.
>
>
> testCheckAndPut(org.apache.hadoop.hbase.rest.client.TestRemoteHTableRetries):
> This is supposed to be overridden by subclasses.
>
>
> testCheckAndDelete(org.apache.hadoop.hbase.rest.client.TestRemoteHTableRetries):
> This is supposed to be overridden by subclasses.
>
>
> testSplitInterval(org.apache.hadoop.hbase.mapreduce.hadoopbackport.TestInputSamplerTool):
> Failed getting constructor
>
>
> testSplitRamdom(org.apache.hadoop.hbase.mapreduce.hadoopbackport.TestInputSamplerTool):
> Failed getting constructor
>
>
> testSplitSample(org.apache.hadoop.hbase.mapreduce.hadoopbackport.TestInputSamplerTool):
> Failed getting constructor
>
>
> testSimpleException(org.apache.hadoop.hbase.errorhandling.TestForeignExceptionSerialization):
> This is supposed to be overridden by subclasses.
>
>
> testRemoteFromLocal(org.apache.hadoop.hbase.errorhandling.TestForeignExceptionSerialization):
> This is supposed to be overridden by subclasses.
>
> Looks like the errors were related to protobuf.
>
> FYI
>
>
>
> On Sat, Jan 11, 2014 at 3:56 AM, Jean-Marc Spaggiari <
> [email protected]> wrote:
>
> > I have also all my servers running with JDK 1.7 for months now without
> > facing any JDK related issue. Even if it's not a real production cluster,
> > it's running MR jobs and clients jobs 24/7.
> >
> >
> > 2014/1/11 lars hofhansl <[email protected]>
> >
> > > Over the course of 0.94.16 I have upgraded junit to 4.11 for 0.94 (this
> > > was done for 0.96 and later a while ago).
> > > With that in place (and a few test fixes, mostly in tests that depended
> > on
> > > the relative ordering in which tests are run)
> > > the testsuite passes on JDK7 as reliable as it does on JDK6.
> > >
> > > I currently have Jenkins builds for both JDK6 and JDK7 that are
> triggered
> > > for each checkin,
> > > I'll keep both running for a while; eventually I am planning to retire
> > the
> > > JDK6 Jenkins.
> > >
> > > At Salesforce we use HBase 0.94 exclusively with JDK7 and have not seen
> > > any issue (at least noen related to JDK7).
> > >
> > > TL;DR: 0.94 should be safe to use with JDK7.
> > >
> > > -- Lars
> >
>

Reply via email to