You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ambari.apache.org by "Sandor Magyari (JIRA)" <ji...@apache.org> on 2017/04/21 13:50:04 UTC

[jira] [Created] (AMBARI-20818) Blueprint processor fails when StackAdvisor recommends unnecessary configuration types

Sandor Magyari created AMBARI-20818:
---------------------------------------

             Summary: Blueprint processor fails when StackAdvisor recommends unnecessary configuration types
                 Key: AMBARI-20818
                 URL: https://issues.apache.org/jira/browse/AMBARI-20818
             Project: Ambari
          Issue Type: Bug
          Components: ambari-server
            Reporter: Sandor Magyari
            Assignee: Sandor Magyari
            Priority: Critical
             Fix For: 3.0.0


When deploying a minimal Blueprint, that only included the Hive MetaStoreServer and Zookeeper, without any other services the deployment failed, and was stuck in the "PENDING" state.  

The root cause is the following exception in ambari-server.log:

{code}
java.lang.NullPointerException
        at org.apache.ambari.server.topology.AmbariContext.waitForConfigurationResolution(AmbariContext.java:430)
        at org.apache.ambari.server.topology.ClusterConfigurationRequest.setConfigurationsOnCluster(ClusterConfigurationRequest.java:496)
        at org.apache.ambari.server.topology.ClusterConfigurationRequest.setConfigurationsOnCluster(ClusterConfigurationRequest.java:406)
        at org.apache.ambari.server.topology.ClusterConfigurationRequest.process(ClusterConfigurationRequest.java:159)
        at org.apache.ambari.server.topology.tasks.
{code}

The problem is that Blueprints processor receives some configuration recommendations from the StackAdvisor for "yarn-site" and "yarn-env", even though Yarn is not present in the cluster. This causes the Blueprint processor to treat these types as updated, and will eventually cause the NullPointerException I mentioned above.

This causes the cluster configuration to be in an inconsistent state, and the Blueprint deployment won't necessarily complete properly, although some manual configuration changes would likely result in a working cluster.

The Blueprint processor should be updated to ignore any configuration types recommended by the StackAdvisor that are not needed for the current cluster.  Usually, these types are from services that are not selected for the cluster.  

If a recommendation is present for a configuration type that is not included in the "desired configurations" for a cluster, the Blueprint processor should ignore this configuration type, and perhaps log some time of warning, so that the user is aware that some recommendations were made that are not being used.  




--
This message was sent by Atlassian JIRA
(v6.3.15#6346)