[jira] [Commented] (AMBARI-17901) Make HDFS operations resilient to namenode safemode

["Andrew Onischuk (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/AMBARI-17901?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16299665#comment-16299665 ] 

Andrew Onischuk commented on AMBARI-17901:
------------------------------------------

Hello [~smagyari]. When starting namenode we now check for safemode. As of fix in AMBARI-17182. All the other operations are done only after successful start of NAMENODE (due to role_command_order.json). Retrying would add some unnecessary delay to operations which mean to fail due to something not working. 

> Make HDFS operations resilient to namenode safemode
> ---------------------------------------------------
>
>                 Key: AMBARI-17901
>                 URL: https://issues.apache.org/jira/browse/AMBARI-17901
>             Project: Ambari
>          Issue Type: Bug
>          Components: ambari-server
>    Affects Versions: 2.4.0
>            Reporter: Magyari Sandor Szilard
>            Assignee: Magyari Sandor Szilard
>             Fix For: 3.0.0
>
>
> HdfsResourceJar and HdfsResourceWebHDFS (WebHDFSUtil) are the classes that carry out the HDFS operations. All retry able operations (e.g. SETPERMISSION) should be guarded with retry logic that would retry the operation until a given timeout before giving up and bailing out.
> To determine which HDFS operations are retry able might be as easy as just looking the returned status/error code or the type of the exception (e.g. "RetriableException") though this needs to be verified if it's consistent with both the webhdfs and hdfsresource jar.
> This problem came up in https://issues.apache.org/jira/browse/AMBARI-17182 when starting all services after Enabling HA. 
> Retry count and timeout should be clarified, as sometimes it may take a long time for namenode to exit safemode.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)