Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
HBase, mail # dev - [VOTE] The 1st hbase-0.96.0 release candidate is available for download


Copy link to this message
-
Re: [VOTE] The 1st hbase-0.96.0 release candidate is available for download
Jean-Marc Spaggiari 2013-09-02, 17:20
Hi St.Ack,

I will open the relate JIRAs in few minutes.

Regarding performances, RandomSeekScanTest is way slower, and only
RandomScanWithRange100Test was faster. Others were similar. For
RandomScanWithRange100Test I suspect that I don't have the right
number for 0.94 so 0.94.11 tests are running right now on the same
server with the same configuration. I will start to have numbers by
end of day, else, tomorrow morning, but will most probably take about
24h to get all of them.

JM

2013/9/2 Stack <[EMAIL PROTECTED]>:
> On Mon, Sep 2, 2013 at 9:41 AM, Jean-Marc Spaggiari <[EMAIL PROTECTED]
>> wrote:
>
>> Got it.
>>
>> I can't run the integration tests for now because I'm lacking some
>> servers :( Need to complete some HBase on RAID tests before I can get
>> those new servers
>>
>> First thing is, start-hbase.cmd has the execute flag set. I don't
>> think it's required. And it will help with tabulation feature if we
>> can un-set it.
>>
>> For 0.96.0RC0 here are my results:
>> First, I get 2 .out files. each time I start the server, instead of
>> usually one... With the same timestamp.
>>
>> -rw-r--r-- 1 jmspaggiari jmspaggiari     0 Aug 31 15:38
>> hbase-jmspaggiari-master-t430s.out
>> -rw-r--r-- 1 jmspaggiari jmspaggiari     0 Aug 31 15:38
>> hbase-jmspaggiari-master-t430s.out.1
>>
>>
>> In the UI, we say "The .META. table holds references to all User Table
>> regions" but the table name is "hbase:meta" and not ".META."
>>
>> On the logs, I found this exception that I did not had before:
>> 2013-08-31 18:45:05,490 WARN
>> [NIOServerCxn.Factory:0.0.0.0/0.0.0.0:2181] server.NIOServerCnxn:
>> caught end of stream exception
>> EndOfStreamException: Unable to read additional data from client
>> sessionid 0x140d68bb9d50004, likely client has closed socket
>>     at
>> org.apache.zookeeper.server.NIOServerCnxn.doIO(NIOServerCnxn.java:220)
>>     at
>> org.apache.zookeeper.server.NIOServerCnxnFactory.run(NIOServerCnxnFactory.java:208)
>>     at java.lang.Thread.run(Thread.java:662)
>>
>>
>> I ran PE over 48h. I don't have another 0.96 baseline to compare with,
>> so I compared with 0.94.
>> org.apache.hadoop.hbase.PerformanceEvaluation$RandomSeekScanTest is
>> about 3 times slower with 0.96.
>> org.apache.hadoop.hbase.PerformanceEvaluation$RandomScanWithRange100Test
>> seems to be 2 times faster.
>>
>> Writes also are faster but I changed my hard drive since I ran 0.94.
>> So I'm currently re-running 0.94 and will provide another more
>> accurate comparison soon.
>>
>> Ran LoadTestTool: Failed to write keys: 0
>>
>> I have not been able to run IntegrationTestLoadAndVerify nor
>> IntegrationTestBigLinkedList because of the lack of servers.
>>
>> So overall, it seems to be working fine, but I have not been able to
>> test this release as deeply as I'm usually testing the 0.94 releases.
>>
>
> Thank you for taking it for a spin JMS -- you the man.  If you don't open
> issues for the teething stuff, I will.
>
> So 0.96 is slower doing RandomSeekScanTest but faster on your other tests?
>  I can take a look too.
>
> St.Ack