[jira] [Created] (AMBARI-8009) Multiple config versions present after Blueprint Cluster Install

["Robert Nettleton (JIRA)" <ji...@apache.org>]

Robert Nettleton created AMBARI-8009:
----------------------------------------

             Summary: Multiple config versions present after Blueprint Cluster Install
                 Key: AMBARI-8009
                 URL: https://issues.apache.org/jira/browse/AMBARI-8009
             Project: Ambari
          Issue Type: Bug
          Components: ambari-server
    Affects Versions: 1.7.0
            Reporter: Robert Nettleton
            Assignee: Robert Nettleton
             Fix For: 1.7.0


After deploying a cluster using Ambari Blueprints, some of the services  (examples: HDFS, Yarn, Hive, etc) will report multiple service configuration versions after the initial startup.  

This is incorrect, since after the first cluster deployment, all configuration should be at Version 1 ("V1").  

The problem occurs because the Ambari Configuration engine has been modified to support versioning on a per-service basis in Ambari overall.  The Blueprint processor currently uses an outdated method to publish the configuration changes prior to a cluster startup, and this is the root of the problem. 

The Blueprint deployment code in ClusterResourceProvider currently publishes a ClusterRequest for each configuration type encountered.  Because each service includes multiple configuration types, the Ambari Configuration framework will increase version number for each type seen for a given service.  

The ClusterResourceProvider needs to be modified to send the ClusterRequest messages at the proper granularity level (one request per service, which includes all config types associated with that service).  

I'm currently working on a patch to resolve this, and will be submitting this sometime soon. 



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)