Fw: Hadoop 2 Namenode HA not working properly

2014-01-24 Thread Bruno Andrade


Begin forwarded message:

Date: Tue, 21 Jan 2014 09:35:23 +
From: Bruno Andrade 
To: user@hadoop.apache.org
Subject: Re: Hadoop 2 Namenode HA not working properly


Hey,

this is my hdfs-site.xml -> http://pastebin.com/qpELkwH8
this is my core-site.xml:


 
 fs.defaultFS
 hdfs://blabla-hadoop
 

 
 hadoop.tmp.dir
 /opt/hadoop/hadoop/tmp
 


I kill only the namenode process and nothing happen, then I killed zkfc 
process and the transition happend, the second namenode become active.

Thanks.



On 01/20/2014 06:44 PM, Jing Zhao wrote:
> Hi Bruno,
>
>  Could you post your configuration? Also, when you killed one of
> the NN, you mean only killing the NN process or you shutdown the whole
> machine?
>
> Thanks,
> -Jing
>
> On Mon, Jan 20, 2014 at 4:11 AM, Bruno Andrade 
> wrote:
>> Hey,
>>
>> I have configured a Hadoop v2.2.0 cluster with QJM and Zookeeper for
>> HA and automatic failover.
>> But I'm having a problem. If I test the automatic failover, by
>> killing one of the namenodes, nothing happens. But if I kill the
>> zkfc of that namenode, then zookeeper elects the other namenode as
>> active.
>>
>> What can it be the problem?
>>
>> Thanks.
>>
>> --
>> Bruno Andrade 
>> Programador (I&D)
>> Eurotux Informática, S.A. | www.eurotux.com
>> (t) +351 253 680 300 (m) +351 936 293 858

-- 
Bruno Andrade 
Programador (I&D)
Eurotux Informática, S.A. | www.eurotux.com
(t) +351 253 680 300 (m) +351 936 293 858



-- 
Bruno Andrade 
Programador (I&D)
Eurotux Informática, S.A. | www.eurotux.com
(t) +351 253 680 300 (m) +351 936 293 858

  
  
Hey,

this is my hdfs-site.xml -> http://pastebin.com/qpELkwH8
this is my core-site.xml:

  
      
          fs.defaultFS
          hdfs://blabla-hadoop
      
  
      
          hadoop.tmp.dir
          /opt/hadoop/hadoop/tmp
      
  

I kill only the namenode process and nothing happen, then I killed
zkfc process and the transition happend, the second namenode become
active.

Thanks.



On 01/20/2014 06:44 PM, Jing Zhao
  wrote:


  Hi Bruno,

Could you post your configuration? Also, when you killed one of
the NN, you mean only killing the NN process or you shutdown the whole
machine?

Thanks,
-Jing

On Mon, Jan 20, 2014 at 4:11 AM, Bruno Andrade  wrote:

  
Hey,

I have configured a Hadoop v2.2.0 cluster with QJM and Zookeeper for HA and
automatic failover.
But I'm having a problem. If I test the automatic failover, by killing one
of the namenodes, nothing happens. But if I kill the zkfc of that namenode,
then zookeeper elects the other namenode as active.

What can it be the problem?

Thanks.

--
Bruno Andrade 
Programador (I&D)
Eurotux Informática, S.A. | www.eurotux.com
(t) +351 253 680 300 (m) +351 936 293 858

  
  



-- 
Bruno Andrade 
Programador (I&D)
Eurotux Informática, S.A. | www.eurotux.com
(t) +351 253 680 300 (m) +351 936 293 858
  



Re: Fw: Hadoop 2 Namenode HA not working properly

2014-01-24 Thread Juan Carlos
Hi Bruno,
ha.zookeeper.quorum is a property in core-site and you have it in
hdfs-site, maybe it's your problem.


2014/1/24 Bruno Andrade 

>
>
> Begin forwarded message:
>
> Date: Tue, 21 Jan 2014 09:35:23 +
> From: Bruno Andrade 
> To: user@hadoop.apache.org
> Subject: Re: Hadoop 2 Namenode HA not working properly
>
>
> Hey,
>
> this is my hdfs-site.xml -> http://pastebin.com/qpELkwH8
> this is my core-site.xml:
>
> 
>  
>  fs.defaultFS
>  hdfs://blabla-hadoop
>  
>
>  
>  hadoop.tmp.dir
>  /opt/hadoop/hadoop/tmp
>  
> 
>
> I kill only the namenode process and nothing happen, then I killed zkfc
> process and the transition happend, the second namenode become active.
>
> Thanks.
>
>
>
> On 01/20/2014 06:44 PM, Jing Zhao wrote:
> > Hi Bruno,
> >
> >  Could you post your configuration? Also, when you killed one of
> > the NN, you mean only killing the NN process or you shutdown the whole
> > machine?
> >
> > Thanks,
> > -Jing
> >
> > On Mon, Jan 20, 2014 at 4:11 AM, Bruno Andrade 
> > wrote:
> >> Hey,
> >>
> >> I have configured a Hadoop v2.2.0 cluster with QJM and Zookeeper for
> >> HA and automatic failover.
> >> But I'm having a problem. If I test the automatic failover, by
> >> killing one of the namenodes, nothing happens. But if I kill the
> >> zkfc of that namenode, then zookeeper elects the other namenode as
> >> active.
> >>
> >> What can it be the problem?
> >>
> >> Thanks.
> >>
> >> --
> >> Bruno Andrade 
> >> Programador (I&D)
> >> Eurotux Informática, S.A. | www.eurotux.com
> >> (t) +351 253 680 300 (m) +351 936 293 858
>
> --
> Bruno Andrade 
> Programador (I&D)
> Eurotux Informática, S.A. | www.eurotux.com
> (t) +351 253 680 300 (m) +351 936 293 858
>
>
>
> --
> Bruno Andrade 
> Programador (I&D)
> Eurotux Informática, S.A. | www.eurotux.com
> (t) +351 253 680 300 (m) +351 936 293 858
>