[ 
https://issues.apache.org/jira/browse/AMBARI-20461?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Di Li updated AMBARI-20461:
---------------------------
    Resolution: Fixed
        Status: Resolved  (was: Patch Available)

> override_uid should set to false when upgrading Ambari 2.1 to 2.2 or newer 
> with custom stacks
> ---------------------------------------------------------------------------------------------
>
>                 Key: AMBARI-20461
>                 URL: https://issues.apache.org/jira/browse/AMBARI-20461
>             Project: Ambari
>          Issue Type: Bug
>          Components: ambari-server
>    Affects Versions: trunk
>            Reporter: Di Li
>            Assignee: Di Li
>             Fix For: trunk
>
>         Attachments: AMBARI-20461.patch
>
>
> Custom stacks may not have override_hbase_uid property in hbase-env.xml, in 
> this case, override_uid is currently set to true by Ambari's auto merge 
> logic. it should be set to false when upgrading Ambari 2.1 to 2.2 or newer 
> with custom stacks in order to respect the existing UID customers already set 
> on their clusters.
> This is for upgrading a third party Ambari/stack distribution from Ambari 
> version 2.1.0 to Ambari trunk. HDP stacks do not have this issue.
> The HBase in my own stack did not have 'override_hbase_uid' when I released 
> it with AMbari 2.1.0. Current code (without my change) will merge 
> override_uid with default value true (defined in cluster-env) when I upgrade 
> Ambari to trunk. For clusters that used UID less than 1000, override_uid=true 
> causes Ambari to update the UIDs, and servers will fail to restart. 
> My code change is in UpgradeCatalog212 and only runs if "override_hbase_uid" 
> does not exist (either it doesn;t exist, or the entire hbase-env doesn;t 
> exist). This means that my code change only affects old Ambari 2.1.0 or 
> Ambari 2.1.1 when upgrading them to Ambari trunk. If a cluster has Ambari 
> 2.2.x (override_uid was added in Ambari 2.1.2), UpgradeCatalog212 does not 
> run, so my logic does not run, respecting whatever override_uid the cluster 
> always uses. If a cluster is with HDP stacks and on Ambari 2.1.0, the current 
> logic in UpgradeCatalog212 runs, so my logic still doesn;t affect it.



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

Reply via email to