-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/33360/#review80791
-----------------------------------------------------------

Ship it!


Ship It!

- Mahadev Konar


On April 20, 2015, 3:04 p.m., Robert Nettleton wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/33360/
> -----------------------------------------------------------
> 
> (Updated April 20, 2015, 3:04 p.m.)
> 
> 
> Review request for Ambari, John Speidel, Mahadev Konar, and Sumit Mohanty.
> 
> 
> Bugs: AMBARI-10487
>     https://issues.apache.org/jira/browse/AMBARI-10487
> 
> 
> Repository: ambari
> 
> 
> Description
> -------
> 
> This patch implements a fix for AMBARI-10487.  
> 
> The BlueprintConfigurationProcessor was incorrectly throwing an exception 
> while processing some HDFS and Yarn properties when an HA cluster was being 
> deployed.  
> 
> The custom handling logic in the config processor was incorrectly throwing an 
> error when more than one instance of an HA-capable service was included in 
> the HA cluster.  In particular, the processor was not accepting FQDN 
> hostnames as valid values for HDFS and Yarn properties.  
> 
> This patch adddresses this problem by:
> 
> 1. Adding handling code in the SingleHostTopologyUpdater to address this 
> issue in HA scenarios for HDFS, Yarn, and Oozie deployments.  
> 2. Modifying the MultipleHostPropertyUpdater to return the original property 
> value if the property does not match any patterns that require hostname 
> substitution. 
> 3. Updating the internal PropertyUpdater interface to pass in the name of the 
> property being processed.  This was done mainly to facilitate simpler 
> debug/error messages, since previously the updater threw an exception without 
> mentioning the actual property involved in the processing failure. 
> 4. Adding unit tests to verify the fixes in this patch.  
> 
> 
> In some future release of Ambari, we should probably build an HA-aware 
> property updater, but at this point it makes more sense to handle this in the 
> existing updaters, in order to support FQDNs as quickly as possible.
> 
> 
> Diffs
> -----
> 
>   
> ambari-server/src/main/java/org/apache/ambari/server/controller/internal/BlueprintConfigurationProcessor.java
>  8247c63 
>   
> ambari-server/src/test/java/org/apache/ambari/server/controller/internal/BlueprintConfigurationProcessorTest.java
>  1839a5f 
> 
> Diff: https://reviews.apache.org/r/33360/diff/
> 
> 
> Testing
> -------
> 
> 1. Ran the ambari-server unit tests.
> 2. Deployed a 3 node HDFS NameNode HA cluster, using a Blueprint that only 
> included FQDN hostnames as property values. 
> 3. Deployed a 3 node HDFS NameNode HA cluster, using a Blueprint that 
> included the usage of the %HOSTGROUP% substitution syntax. 
> 4. Deployed a 3 node Yarn ResourceManager HA cluster, using a Blueprint that 
> only included FQDN hostnames as property values.
> 5. Deployed a 3 node Yarn ResourceManager HA cluster, using a Blueprint that 
> included the usage of the %HOSTGROUP% substitution syntax in certain property 
> values (HOSTGROUP support for remaining Yarn properties to be added in a 
> subsequent patch).
> 
> 
> Thanks,
> 
> Robert Nettleton
> 
>

Reply via email to