[
https://issues.apache.org/jira/browse/IGNITE-345?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14377526#comment-14377526
]
Dmitriy Setrakyan commented on IGNITE-345:
------------------------------------------
Vladimir, I have not looked into all the details of this ticket, but can you
please make sure that IGFS is properly documented on readme.io?
> documentation: better describe secondary filesystem configuration
> -----------------------------------------------------------------
>
> Key: IGNITE-345
> URL: https://issues.apache.org/jira/browse/IGNITE-345
> Project: Ignite
> Issue Type: Wish
> Components: documentation, hadoop
> Affects Versions: sprint-1
> Reporter: Ivan Veselovsky
> Assignee: Vladimir Ozerov
> Priority: Critical
> Fix For: sprint-3
>
>
> Currently the documentation on how to configure the secondary HDFS
> (http://hadoop.gridgain.org/latest/GGFS) is not quite clear.
> 1) The key point there is that this mode requires two different Hadoop
> configurations (and, perhaps, installations): Hadoop server should have
> ordinary hdfs filesystem configured, like hdfs://aaa:9000/ , while Hadoop
> client should use igfs:// filesystem , that, in turn, runs upon the hdfs
> filesystem configured in the Hadoop server through the Ignite node(s). This
> way, the Hadoop client and Hadoop server have its own core-site.xml each. In
> case of local deployment (everything on one machine) this may be done as 2
> different Hadoop configurations or 2 different copies of Hadoop
> installation. We'd better provide comprehensive examples of the configuration
> in the documentation.
> 2) the second point that is quite unclear is the purpose of "cfgPath"
> constructor attribute (see below) of the Ignite node configuration. Actually
> it supposed to mean the path to the *server* Hadoop configuration
> (core-site.xml), but actually it is optional (currently it is impossible to
> omit this paramater, but it is possible to use empty config file.) This
> config is used (1) to get the secondary hdfs URI, if the one is not given
> directly as "uri" paramater , and (2) to get additional parameters of that
> filesystem, if any. The difficulty there is that this config file may (and
> should?) actually reside on a different node (host), so it may be tricky to
> provide local URL to this configuration.
> {code}
> <property name="igfsConfiguration">
> <list>
> <bean
> class="org.apache.ignite.configuration.IgfsConfiguration"
> parent="igfsCfgBase">
> <property name="name" value="igfs"/>
> ....
> <property name="defaultMode" value="PROXY"/>
> <property name="secondaryFileSystem">
> <bean
> class="org.apache.ignite.internal.igfs.hadoop.IgfsHadoopFileSystemWrapper">
> <constructor-arg name="uri"
> value="hdfs://172.17.0.4:9000"/>
> <constructor-arg name="cfgPath"
> value="....../etc/hadoop/core-site.xml"/>
> </bean>
> </property>
> </bean>
> </list>
> </property>
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
