You are viewing a plain text version of this content. The canonical link for it is here.
Posted to hdfs-user@hadoop.apache.org by ESGLinux <es...@gmail.com> on 2012/12/21 11:48:38 UTC

understanding federation and mount points

Hi All,

I´m newbie with federation (thanks to Harsh J. for his help with the nodes
;-) ) and I have a doubt about how clients mount the namespaces:

If I have a NS1 that stores data for the users, for example /datausers
And I have a NS2 that stores  apache logs, for example /apachelogs

I suposse I need to configure the clients something like this:

<property>
   <name>fs.viewfs.mounttable.cluster.link./datausers</name>
   <value>hdfs://namenode01.8020/datausers</value>
</property>
<property>
   <name>fs.viewfs.mounttable.cluster.link./apachelogs</name>
   <value>hdfs://namenode02:8020/apachelogs</value>
</property>

My question is what happens when a client try to put a file in a directorio
at top level,  for example /otherDir/. Who handles / dir ?

I have read that it´s similar to fstab but in linux you always have /
mounted in a device, it´s necessary to do the same with federation?

Thank in advance,

ESGLinux,

Re: understanding federation and mount points

Posted by ESGLinux <es...@gmail.com>.
Hi All,

looking for information to resolve my doubt I have found this interesting
video:
http://www.cloudera.com/content/cloudera/en/resources/library/training/whats-new-in-cdh4-a-guide-for-previous-attendees-of-cloudera-administrator-training-for-apache-hadoop.html

In the minute 33 the video resolves it:

When a client tries to access a path not handled by a namenode it receives
an exception,

Question solved!!!

Greetings and Merry Christmas!!!

ESGLinux



2012/12/21 ESGLinux <es...@gmail.com>

> Hi All,
>
> I´m newbie with federation (thanks to Harsh J. for his help with the nodes
> ;-) ) and I have a doubt about how clients mount the namespaces:
>
> If I have a NS1 that stores data for the users, for example /datausers
> And I have a NS2 that stores  apache logs, for example /apachelogs
>
> I suposse I need to configure the clients something like this:
>
> <property>
>    <name>fs.viewfs.mounttable.cluster.link./datausers</name>
>    <value>hdfs://namenode01.8020/datausers</value>
> </property>
> <property>
>    <name>fs.viewfs.mounttable.cluster.link./apachelogs</name>
>    <value>hdfs://namenode02:8020/apachelogs</value>
> </property>
>
> My question is what happens when a client try to put a file in a
> directorio at top level,  for example /otherDir/. Who handles / dir ?
>
> I have read that it´s similar to fstab but in linux you always have /
> mounted in a device, it´s necessary to do the same with federation?
>
> Thank in advance,
>
> ESGLinux,
>
>
>
>
>
>
>

Re: understanding federation and mount points

Posted by ESGLinux <es...@gmail.com>.
Hi All,

looking for information to resolve my doubt I have found this interesting
video:
http://www.cloudera.com/content/cloudera/en/resources/library/training/whats-new-in-cdh4-a-guide-for-previous-attendees-of-cloudera-administrator-training-for-apache-hadoop.html

In the minute 33 the video resolves it:

When a client tries to access a path not handled by a namenode it receives
an exception,

Question solved!!!

Greetings and Merry Christmas!!!

ESGLinux



2012/12/21 ESGLinux <es...@gmail.com>

> Hi All,
>
> I´m newbie with federation (thanks to Harsh J. for his help with the nodes
> ;-) ) and I have a doubt about how clients mount the namespaces:
>
> If I have a NS1 that stores data for the users, for example /datausers
> And I have a NS2 that stores  apache logs, for example /apachelogs
>
> I suposse I need to configure the clients something like this:
>
> <property>
>    <name>fs.viewfs.mounttable.cluster.link./datausers</name>
>    <value>hdfs://namenode01.8020/datausers</value>
> </property>
> <property>
>    <name>fs.viewfs.mounttable.cluster.link./apachelogs</name>
>    <value>hdfs://namenode02:8020/apachelogs</value>
> </property>
>
> My question is what happens when a client try to put a file in a
> directorio at top level,  for example /otherDir/. Who handles / dir ?
>
> I have read that it´s similar to fstab but in linux you always have /
> mounted in a device, it´s necessary to do the same with federation?
>
> Thank in advance,
>
> ESGLinux,
>
>
>
>
>
>
>

Re: understanding federation and mount points

Posted by ESGLinux <es...@gmail.com>.
Hi All,

looking for information to resolve my doubt I have found this interesting
video:
http://www.cloudera.com/content/cloudera/en/resources/library/training/whats-new-in-cdh4-a-guide-for-previous-attendees-of-cloudera-administrator-training-for-apache-hadoop.html

In the minute 33 the video resolves it:

When a client tries to access a path not handled by a namenode it receives
an exception,

Question solved!!!

Greetings and Merry Christmas!!!

ESGLinux



2012/12/21 ESGLinux <es...@gmail.com>

> Hi All,
>
> I´m newbie with federation (thanks to Harsh J. for his help with the nodes
> ;-) ) and I have a doubt about how clients mount the namespaces:
>
> If I have a NS1 that stores data for the users, for example /datausers
> And I have a NS2 that stores  apache logs, for example /apachelogs
>
> I suposse I need to configure the clients something like this:
>
> <property>
>    <name>fs.viewfs.mounttable.cluster.link./datausers</name>
>    <value>hdfs://namenode01.8020/datausers</value>
> </property>
> <property>
>    <name>fs.viewfs.mounttable.cluster.link./apachelogs</name>
>    <value>hdfs://namenode02:8020/apachelogs</value>
> </property>
>
> My question is what happens when a client try to put a file in a
> directorio at top level,  for example /otherDir/. Who handles / dir ?
>
> I have read that it´s similar to fstab but in linux you always have /
> mounted in a device, it´s necessary to do the same with federation?
>
> Thank in advance,
>
> ESGLinux,
>
>
>
>
>
>
>

Re: understanding federation and mount points

Posted by ESGLinux <es...@gmail.com>.
Hi All,

looking for information to resolve my doubt I have found this interesting
video:
http://www.cloudera.com/content/cloudera/en/resources/library/training/whats-new-in-cdh4-a-guide-for-previous-attendees-of-cloudera-administrator-training-for-apache-hadoop.html

In the minute 33 the video resolves it:

When a client tries to access a path not handled by a namenode it receives
an exception,

Question solved!!!

Greetings and Merry Christmas!!!

ESGLinux



2012/12/21 ESGLinux <es...@gmail.com>

> Hi All,
>
> I´m newbie with federation (thanks to Harsh J. for his help with the nodes
> ;-) ) and I have a doubt about how clients mount the namespaces:
>
> If I have a NS1 that stores data for the users, for example /datausers
> And I have a NS2 that stores  apache logs, for example /apachelogs
>
> I suposse I need to configure the clients something like this:
>
> <property>
>    <name>fs.viewfs.mounttable.cluster.link./datausers</name>
>    <value>hdfs://namenode01.8020/datausers</value>
> </property>
> <property>
>    <name>fs.viewfs.mounttable.cluster.link./apachelogs</name>
>    <value>hdfs://namenode02:8020/apachelogs</value>
> </property>
>
> My question is what happens when a client try to put a file in a
> directorio at top level,  for example /otherDir/. Who handles / dir ?
>
> I have read that it´s similar to fstab but in linux you always have /
> mounted in a device, it´s necessary to do the same with federation?
>
> Thank in advance,
>
> ESGLinux,
>
>
>
>
>
>
>