You are viewing a plain text version of this content. The canonical link for it is here.
Posted to mapreduce-user@hadoop.apache.org by Tomasz Chmielewski <ma...@wpkg.org> on 2013/09/03 22:54:50 UTC
so the master just died... now what?
Just starting with hadoop and hbase, and couldn't find specific answers
in official documentation (unless I've missed the obvious).
Assuming I have three hadoop servers: h1, h2 and h3, with h1 being a
master+slave - what is a recovery scenario if the master server, h1,
died and is beyond repair (burned with all disks and got flooded)?
Do I just edit conf/masters file on any of the remaining slaves (say,
h2), make it a master, and start the NameNode and JobTracker there?
Can anyone point me to relevant documentation?
--
Tomasz Chmielewski
http://wpkg.org
Re: so the master just died... now what?
Posted by Mohammad Tariq <do...@gmail.com>.
Hello Tomasz,
Just to add,
Although it says *masters*, the */conf/masters* actually specifies the
machine where *SecondaryNameNode* will run. Master daemons run on the
machine where you execute the start scripts. If you need to change the
master machine, you must make appropriate changes in the *core-site.xml *and
*mapred-site.xml* files. Also, update the IP and hostname in the *
/etc/hosts* file of your slaves.
Warm Regards,
Tariq
cloudfront.blogspot.com
On Wed, Sep 4, 2013 at 2:31 AM, Shahab Yunus <sh...@gmail.com> wrote:
>
> Keep in mind that there are 2 flavors of Hadoop: the older one without HA
> and the new one with it. Anyway, have you seen the following?
>
> http://wiki.apache.org/hadoop/NameNodeFailover
> http://www.youtube.com/watch?v=Ln1GMkQvP9w
>
> http://docs.hortonworks.com/HDPDocuments/HDP1/HDP-1.3.0/bk_hdp1-system-admin-guide/content/sysadminguides_ha_chap2_5_5.html
>
> Regards,
> Shahab
>
>
> On Tue, Sep 3, 2013 at 4:54 PM, Tomasz Chmielewski <ma...@wpkg.org>wrote:
>
>> Just starting with hadoop and hbase, and couldn't find specific answers
>> in official documentation (unless I've missed the obvious).
>>
>>
>> Assuming I have three hadoop servers: h1, h2 and h3, with h1 being a
>> master+slave - what is a recovery scenario if the master server, h1,
>> died and is beyond repair (burned with all disks and got flooded)?
>>
>> Do I just edit conf/masters file on any of the remaining slaves (say,
>> h2), make it a master, and start the NameNode and JobTracker there?
>>
>> Can anyone point me to relevant documentation?
>>
>>
>> --
>> Tomasz Chmielewski
>> http://wpkg.org
>>
>>
>
Re: so the master just died... now what?
Posted by Mohammad Tariq <do...@gmail.com>.
Hello Tomasz,
Just to add,
Although it says *masters*, the */conf/masters* actually specifies the
machine where *SecondaryNameNode* will run. Master daemons run on the
machine where you execute the start scripts. If you need to change the
master machine, you must make appropriate changes in the *core-site.xml *and
*mapred-site.xml* files. Also, update the IP and hostname in the *
/etc/hosts* file of your slaves.
Warm Regards,
Tariq
cloudfront.blogspot.com
On Wed, Sep 4, 2013 at 2:31 AM, Shahab Yunus <sh...@gmail.com> wrote:
>
> Keep in mind that there are 2 flavors of Hadoop: the older one without HA
> and the new one with it. Anyway, have you seen the following?
>
> http://wiki.apache.org/hadoop/NameNodeFailover
> http://www.youtube.com/watch?v=Ln1GMkQvP9w
>
> http://docs.hortonworks.com/HDPDocuments/HDP1/HDP-1.3.0/bk_hdp1-system-admin-guide/content/sysadminguides_ha_chap2_5_5.html
>
> Regards,
> Shahab
>
>
> On Tue, Sep 3, 2013 at 4:54 PM, Tomasz Chmielewski <ma...@wpkg.org>wrote:
>
>> Just starting with hadoop and hbase, and couldn't find specific answers
>> in official documentation (unless I've missed the obvious).
>>
>>
>> Assuming I have three hadoop servers: h1, h2 and h3, with h1 being a
>> master+slave - what is a recovery scenario if the master server, h1,
>> died and is beyond repair (burned with all disks and got flooded)?
>>
>> Do I just edit conf/masters file on any of the remaining slaves (say,
>> h2), make it a master, and start the NameNode and JobTracker there?
>>
>> Can anyone point me to relevant documentation?
>>
>>
>> --
>> Tomasz Chmielewski
>> http://wpkg.org
>>
>>
>
Re: so the master just died... now what?
Posted by Mohammad Tariq <do...@gmail.com>.
Hello Tomasz,
Just to add,
Although it says *masters*, the */conf/masters* actually specifies the
machine where *SecondaryNameNode* will run. Master daemons run on the
machine where you execute the start scripts. If you need to change the
master machine, you must make appropriate changes in the *core-site.xml *and
*mapred-site.xml* files. Also, update the IP and hostname in the *
/etc/hosts* file of your slaves.
Warm Regards,
Tariq
cloudfront.blogspot.com
On Wed, Sep 4, 2013 at 2:31 AM, Shahab Yunus <sh...@gmail.com> wrote:
>
> Keep in mind that there are 2 flavors of Hadoop: the older one without HA
> and the new one with it. Anyway, have you seen the following?
>
> http://wiki.apache.org/hadoop/NameNodeFailover
> http://www.youtube.com/watch?v=Ln1GMkQvP9w
>
> http://docs.hortonworks.com/HDPDocuments/HDP1/HDP-1.3.0/bk_hdp1-system-admin-guide/content/sysadminguides_ha_chap2_5_5.html
>
> Regards,
> Shahab
>
>
> On Tue, Sep 3, 2013 at 4:54 PM, Tomasz Chmielewski <ma...@wpkg.org>wrote:
>
>> Just starting with hadoop and hbase, and couldn't find specific answers
>> in official documentation (unless I've missed the obvious).
>>
>>
>> Assuming I have three hadoop servers: h1, h2 and h3, with h1 being a
>> master+slave - what is a recovery scenario if the master server, h1,
>> died and is beyond repair (burned with all disks and got flooded)?
>>
>> Do I just edit conf/masters file on any of the remaining slaves (say,
>> h2), make it a master, and start the NameNode and JobTracker there?
>>
>> Can anyone point me to relevant documentation?
>>
>>
>> --
>> Tomasz Chmielewski
>> http://wpkg.org
>>
>>
>
Re: so the master just died... now what?
Posted by Tomasz Chmielewski <ma...@wpkg.org>.
I'm using Hadoop 1.1.2, since the latest stable HBase (0.94.x) does not
seem to be supported when using Hadoop 2.x (or, at least, was not tested
well enough).
So what are my best options?
Use Hadoop 2.x after all, and follow one of these?
http://hadoop.apache.org/docs/current/hadoop-yarn/hadoop-yarn-site/HDFSHighAvailabilityWithQJM.html
http://hadoop.apache.org/docs/current/hadoop-yarn/hadoop-yarn-site/HDFSHighAvailabilityWithNFS.html
Or continue using Hadoop 1.1.2?
--
Tomasz Chmielewski
http://wpkg.org
On 09/04/2013 06:01 AM, Shahab Yunus wrote:
>
> Keep in mind that there are 2 flavors of Hadoop: the older one without
> HA and the new one with it. Anyway, have you seen the following?
>
> http://wiki.apache.org/hadoop/NameNodeFailover
> http://www.youtube.com/watch?v=Ln1GMkQvP9w
> http://docs.hortonworks.com/HDPDocuments/HDP1/HDP-1.3.0/bk_hdp1-system-admin-guide/content/sysadminguides_ha_chap2_5_5.html
>
> Regards,
> Shahab
>
>
> On Tue, Sep 3, 2013 at 4:54 PM, Tomasz Chmielewski <mangoo@wpkg.org
> <ma...@wpkg.org>> wrote:
>
> Just starting with hadoop and hbase, and couldn't find specific answers
> in official documentation (unless I've missed the obvious).
>
>
> Assuming I have three hadoop servers: h1, h2 and h3, with h1 being a
> master+slave - what is a recovery scenario if the master server, h1,
> died and is beyond repair (burned with all disks and got flooded)?
>
> Do I just edit conf/masters file on any of the remaining slaves (say,
> h2), make it a master, and start the NameNode and JobTracker there?
>
> Can anyone point me to relevant documentation?
>
>
> --
> Tomasz Chmielewski
> http://wpkg.org
>
>
Re: so the master just died... now what?
Posted by Tomasz Chmielewski <ma...@wpkg.org>.
I'm using Hadoop 1.1.2, since the latest stable HBase (0.94.x) does not
seem to be supported when using Hadoop 2.x (or, at least, was not tested
well enough).
So what are my best options?
Use Hadoop 2.x after all, and follow one of these?
http://hadoop.apache.org/docs/current/hadoop-yarn/hadoop-yarn-site/HDFSHighAvailabilityWithQJM.html
http://hadoop.apache.org/docs/current/hadoop-yarn/hadoop-yarn-site/HDFSHighAvailabilityWithNFS.html
Or continue using Hadoop 1.1.2?
--
Tomasz Chmielewski
http://wpkg.org
On 09/04/2013 06:01 AM, Shahab Yunus wrote:
>
> Keep in mind that there are 2 flavors of Hadoop: the older one without
> HA and the new one with it. Anyway, have you seen the following?
>
> http://wiki.apache.org/hadoop/NameNodeFailover
> http://www.youtube.com/watch?v=Ln1GMkQvP9w
> http://docs.hortonworks.com/HDPDocuments/HDP1/HDP-1.3.0/bk_hdp1-system-admin-guide/content/sysadminguides_ha_chap2_5_5.html
>
> Regards,
> Shahab
>
>
> On Tue, Sep 3, 2013 at 4:54 PM, Tomasz Chmielewski <mangoo@wpkg.org
> <ma...@wpkg.org>> wrote:
>
> Just starting with hadoop and hbase, and couldn't find specific answers
> in official documentation (unless I've missed the obvious).
>
>
> Assuming I have three hadoop servers: h1, h2 and h3, with h1 being a
> master+slave - what is a recovery scenario if the master server, h1,
> died and is beyond repair (burned with all disks and got flooded)?
>
> Do I just edit conf/masters file on any of the remaining slaves (say,
> h2), make it a master, and start the NameNode and JobTracker there?
>
> Can anyone point me to relevant documentation?
>
>
> --
> Tomasz Chmielewski
> http://wpkg.org
>
>
Re: so the master just died... now what?
Posted by Mohammad Tariq <do...@gmail.com>.
Hello Tomasz,
Just to add,
Although it says *masters*, the */conf/masters* actually specifies the
machine where *SecondaryNameNode* will run. Master daemons run on the
machine where you execute the start scripts. If you need to change the
master machine, you must make appropriate changes in the *core-site.xml *and
*mapred-site.xml* files. Also, update the IP and hostname in the *
/etc/hosts* file of your slaves.
Warm Regards,
Tariq
cloudfront.blogspot.com
On Wed, Sep 4, 2013 at 2:31 AM, Shahab Yunus <sh...@gmail.com> wrote:
>
> Keep in mind that there are 2 flavors of Hadoop: the older one without HA
> and the new one with it. Anyway, have you seen the following?
>
> http://wiki.apache.org/hadoop/NameNodeFailover
> http://www.youtube.com/watch?v=Ln1GMkQvP9w
>
> http://docs.hortonworks.com/HDPDocuments/HDP1/HDP-1.3.0/bk_hdp1-system-admin-guide/content/sysadminguides_ha_chap2_5_5.html
>
> Regards,
> Shahab
>
>
> On Tue, Sep 3, 2013 at 4:54 PM, Tomasz Chmielewski <ma...@wpkg.org>wrote:
>
>> Just starting with hadoop and hbase, and couldn't find specific answers
>> in official documentation (unless I've missed the obvious).
>>
>>
>> Assuming I have three hadoop servers: h1, h2 and h3, with h1 being a
>> master+slave - what is a recovery scenario if the master server, h1,
>> died and is beyond repair (burned with all disks and got flooded)?
>>
>> Do I just edit conf/masters file on any of the remaining slaves (say,
>> h2), make it a master, and start the NameNode and JobTracker there?
>>
>> Can anyone point me to relevant documentation?
>>
>>
>> --
>> Tomasz Chmielewski
>> http://wpkg.org
>>
>>
>
Re: so the master just died... now what?
Posted by Tomasz Chmielewski <ma...@wpkg.org>.
I'm using Hadoop 1.1.2, since the latest stable HBase (0.94.x) does not
seem to be supported when using Hadoop 2.x (or, at least, was not tested
well enough).
So what are my best options?
Use Hadoop 2.x after all, and follow one of these?
http://hadoop.apache.org/docs/current/hadoop-yarn/hadoop-yarn-site/HDFSHighAvailabilityWithQJM.html
http://hadoop.apache.org/docs/current/hadoop-yarn/hadoop-yarn-site/HDFSHighAvailabilityWithNFS.html
Or continue using Hadoop 1.1.2?
--
Tomasz Chmielewski
http://wpkg.org
On 09/04/2013 06:01 AM, Shahab Yunus wrote:
>
> Keep in mind that there are 2 flavors of Hadoop: the older one without
> HA and the new one with it. Anyway, have you seen the following?
>
> http://wiki.apache.org/hadoop/NameNodeFailover
> http://www.youtube.com/watch?v=Ln1GMkQvP9w
> http://docs.hortonworks.com/HDPDocuments/HDP1/HDP-1.3.0/bk_hdp1-system-admin-guide/content/sysadminguides_ha_chap2_5_5.html
>
> Regards,
> Shahab
>
>
> On Tue, Sep 3, 2013 at 4:54 PM, Tomasz Chmielewski <mangoo@wpkg.org
> <ma...@wpkg.org>> wrote:
>
> Just starting with hadoop and hbase, and couldn't find specific answers
> in official documentation (unless I've missed the obvious).
>
>
> Assuming I have three hadoop servers: h1, h2 and h3, with h1 being a
> master+slave - what is a recovery scenario if the master server, h1,
> died and is beyond repair (burned with all disks and got flooded)?
>
> Do I just edit conf/masters file on any of the remaining slaves (say,
> h2), make it a master, and start the NameNode and JobTracker there?
>
> Can anyone point me to relevant documentation?
>
>
> --
> Tomasz Chmielewski
> http://wpkg.org
>
>
Re: so the master just died... now what?
Posted by Tomasz Chmielewski <ma...@wpkg.org>.
I'm using Hadoop 1.1.2, since the latest stable HBase (0.94.x) does not
seem to be supported when using Hadoop 2.x (or, at least, was not tested
well enough).
So what are my best options?
Use Hadoop 2.x after all, and follow one of these?
http://hadoop.apache.org/docs/current/hadoop-yarn/hadoop-yarn-site/HDFSHighAvailabilityWithQJM.html
http://hadoop.apache.org/docs/current/hadoop-yarn/hadoop-yarn-site/HDFSHighAvailabilityWithNFS.html
Or continue using Hadoop 1.1.2?
--
Tomasz Chmielewski
http://wpkg.org
On 09/04/2013 06:01 AM, Shahab Yunus wrote:
>
> Keep in mind that there are 2 flavors of Hadoop: the older one without
> HA and the new one with it. Anyway, have you seen the following?
>
> http://wiki.apache.org/hadoop/NameNodeFailover
> http://www.youtube.com/watch?v=Ln1GMkQvP9w
> http://docs.hortonworks.com/HDPDocuments/HDP1/HDP-1.3.0/bk_hdp1-system-admin-guide/content/sysadminguides_ha_chap2_5_5.html
>
> Regards,
> Shahab
>
>
> On Tue, Sep 3, 2013 at 4:54 PM, Tomasz Chmielewski <mangoo@wpkg.org
> <ma...@wpkg.org>> wrote:
>
> Just starting with hadoop and hbase, and couldn't find specific answers
> in official documentation (unless I've missed the obvious).
>
>
> Assuming I have three hadoop servers: h1, h2 and h3, with h1 being a
> master+slave - what is a recovery scenario if the master server, h1,
> died and is beyond repair (burned with all disks and got flooded)?
>
> Do I just edit conf/masters file on any of the remaining slaves (say,
> h2), make it a master, and start the NameNode and JobTracker there?
>
> Can anyone point me to relevant documentation?
>
>
> --
> Tomasz Chmielewski
> http://wpkg.org
>
>
Re: so the master just died... now what?
Posted by Shahab Yunus <sh...@gmail.com>.
Keep in mind that there are 2 flavors of Hadoop: the older one without HA
and the new one with it. Anyway, have you seen the following?
http://wiki.apache.org/hadoop/NameNodeFailover
http://www.youtube.com/watch?v=Ln1GMkQvP9w
http://docs.hortonworks.com/HDPDocuments/HDP1/HDP-1.3.0/bk_hdp1-system-admin-guide/content/sysadminguides_ha_chap2_5_5.html
Regards,
Shahab
On Tue, Sep 3, 2013 at 4:54 PM, Tomasz Chmielewski <ma...@wpkg.org> wrote:
> Just starting with hadoop and hbase, and couldn't find specific answers
> in official documentation (unless I've missed the obvious).
>
>
> Assuming I have three hadoop servers: h1, h2 and h3, with h1 being a
> master+slave - what is a recovery scenario if the master server, h1,
> died and is beyond repair (burned with all disks and got flooded)?
>
> Do I just edit conf/masters file on any of the remaining slaves (say,
> h2), make it a master, and start the NameNode and JobTracker there?
>
> Can anyone point me to relevant documentation?
>
>
> --
> Tomasz Chmielewski
> http://wpkg.org
>
>
Re: so the master just died... now what?
Posted by Shahab Yunus <sh...@gmail.com>.
Keep in mind that there are 2 flavors of Hadoop: the older one without HA
and the new one with it. Anyway, have you seen the following?
http://wiki.apache.org/hadoop/NameNodeFailover
http://www.youtube.com/watch?v=Ln1GMkQvP9w
http://docs.hortonworks.com/HDPDocuments/HDP1/HDP-1.3.0/bk_hdp1-system-admin-guide/content/sysadminguides_ha_chap2_5_5.html
Regards,
Shahab
On Tue, Sep 3, 2013 at 4:54 PM, Tomasz Chmielewski <ma...@wpkg.org> wrote:
> Just starting with hadoop and hbase, and couldn't find specific answers
> in official documentation (unless I've missed the obvious).
>
>
> Assuming I have three hadoop servers: h1, h2 and h3, with h1 being a
> master+slave - what is a recovery scenario if the master server, h1,
> died and is beyond repair (burned with all disks and got flooded)?
>
> Do I just edit conf/masters file on any of the remaining slaves (say,
> h2), make it a master, and start the NameNode and JobTracker there?
>
> Can anyone point me to relevant documentation?
>
>
> --
> Tomasz Chmielewski
> http://wpkg.org
>
>
Re: so the master just died... now what?
Posted by Shahab Yunus <sh...@gmail.com>.
Keep in mind that there are 2 flavors of Hadoop: the older one without HA
and the new one with it. Anyway, have you seen the following?
http://wiki.apache.org/hadoop/NameNodeFailover
http://www.youtube.com/watch?v=Ln1GMkQvP9w
http://docs.hortonworks.com/HDPDocuments/HDP1/HDP-1.3.0/bk_hdp1-system-admin-guide/content/sysadminguides_ha_chap2_5_5.html
Regards,
Shahab
On Tue, Sep 3, 2013 at 4:54 PM, Tomasz Chmielewski <ma...@wpkg.org> wrote:
> Just starting with hadoop and hbase, and couldn't find specific answers
> in official documentation (unless I've missed the obvious).
>
>
> Assuming I have three hadoop servers: h1, h2 and h3, with h1 being a
> master+slave - what is a recovery scenario if the master server, h1,
> died and is beyond repair (burned with all disks and got flooded)?
>
> Do I just edit conf/masters file on any of the remaining slaves (say,
> h2), make it a master, and start the NameNode and JobTracker there?
>
> Can anyone point me to relevant documentation?
>
>
> --
> Tomasz Chmielewski
> http://wpkg.org
>
>
Re: so the master just died... now what?
Posted by Shahab Yunus <sh...@gmail.com>.
Keep in mind that there are 2 flavors of Hadoop: the older one without HA
and the new one with it. Anyway, have you seen the following?
http://wiki.apache.org/hadoop/NameNodeFailover
http://www.youtube.com/watch?v=Ln1GMkQvP9w
http://docs.hortonworks.com/HDPDocuments/HDP1/HDP-1.3.0/bk_hdp1-system-admin-guide/content/sysadminguides_ha_chap2_5_5.html
Regards,
Shahab
On Tue, Sep 3, 2013 at 4:54 PM, Tomasz Chmielewski <ma...@wpkg.org> wrote:
> Just starting with hadoop and hbase, and couldn't find specific answers
> in official documentation (unless I've missed the obvious).
>
>
> Assuming I have three hadoop servers: h1, h2 and h3, with h1 being a
> master+slave - what is a recovery scenario if the master server, h1,
> died and is beyond repair (burned with all disks and got flooded)?
>
> Do I just edit conf/masters file on any of the remaining slaves (say,
> h2), make it a master, and start the NameNode and JobTracker there?
>
> Can anyone point me to relevant documentation?
>
>
> --
> Tomasz Chmielewski
> http://wpkg.org
>
>