> On 2012-03-06 10:13:38, Mohammad Islam wrote: > > Hi Alejandro, > > So how/who will the hadoop-config.xml be populated? Because it is expected > > to be cluster specific configurations defined in *-site.xml (such as core, > > hdfs etc). > > Is it we have to create another file? Or we can **reuse** the cluster > > specified *-site.xml as it is. Only thing we have to configure those file > > names into oozie-site.xml. > > right? > > > > Regards, > > Mohammad > > > > > > Alejandro Abdelnur wrote: > Mohammad, > > you have to create a hadoop-config.xml file per cluster, as you indicated > typically will have the information the *-site.xml files of the cluster used > by the clients. You could reuse the cluster specific ones by renaming them, > still you may have to combine the core/hdfs/mapred/yarn. And yes, the only > thing is to configure them in the the oozie-site.xml and copy them into the > oozie conf/ dir. > > Mohammad Islam wrote: > can't we use them by reusing the file as it is without renaming and > without combining? > This is another level of task and indirection. > Renaming + combining could be handled through some configuration in > oozie-site.xml. > Because there will be an impact on us. > > >
Mohammad, Are you suggesting that instead pointing to a file we should point to a directory and from there load all the hadoop *-site.xml files ? We could do that, yes it would be simpler for admins. My concern here is that we would have to be aware from Oozie side on the order of loading the *-site.xml files. That is why I've thought not to do it. But you have point on usability. Thoughts? - Alejandro ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/4186/#review5637 ----------------------------------------------------------- On 2012-03-06 05:39:54, Alejandro Abdelnur wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/4186/ > ----------------------------------------------------------- > > (Updated 2012-03-06 05:39:54) > > > Review request for oozie. > > > Summary > ------- > > hadoop.configurations= [<AUTHORITY>=<CONFIG_FILE>]* > > Where <AUTHORITY> is the HOST:PORT of the Hadoop JT/NN and <CONFIG_FILE> is > the Hadoop configuration file to load from the local filesystem with the > defaults for the corresponding JT/NN. > > In addition, a '*' wildcard would be support for the <AUTHORITY>, this would > be used as default if the job used authority does not have a custom config. > > The Hadoop configuration file would be looked in the Oozie conf/ directory > > An example of this configuration would be: > > hadoop.configuration= *=hadoop-defaultconf.xml > myjobtracker:8021=hadoop-cluster1.xml, mynamenode:8020=hadoop-cluster1.xml > > Documentation is still missing, coming soon. > > > This addresses bug OOZIE-736. > https://issues.apache.org/jira/browse/OOZIE-736 > > > Diffs > ----- > > trunk/core/src/main/conf/hadoop-config.xml PRE-CREATION > > trunk/core/src/main/java/org/apache/oozie/service/HadoopAccessorService.java > 1297338 > trunk/core/src/main/resources/oozie-default.xml 1297338 > > trunk/core/src/test/java/org/apache/oozie/service/TestHadoopAccessorService.java > 1297338 > trunk/core/src/test/java/org/apache/oozie/test/XFsTestCase.java 1297338 > trunk/core/src/test/java/org/apache/oozie/test/XTestCase.java 1297338 > trunk/core/src/test/resources/hadoop-config.xml PRE-CREATION > trunk/core/src/test/resources/test-hadoop-config.xml PRE-CREATION > trunk/docs/src/site/twiki/AG_HadoopConfiguration.twiki PRE-CREATION > trunk/docs/src/site/twiki/DG_UsingHadoopKerberos.twiki 1297338 > trunk/docs/src/site/twiki/index.twiki 1297338 > > Diff: https://reviews.apache.org/r/4186/diff > > > Testing > ------- > > run all testcases with Hadoop 1 > > > Thanks, > > Alejandro > >
