Xuan Gong updated YARN-4101:
    Attachment: YARN-4101.2.patch

Fix the testcase failure

> RM should print alert messages if Zookeeper and Resourcemanager gets 
> connection issue
> -------------------------------------------------------------------------------------
>                 Key: YARN-4101
>                 URL: https://issues.apache.org/jira/browse/YARN-4101
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: yarn
>            Reporter: Yesha Vora
>            Assignee: Xuan Gong
>            Priority: Critical
>         Attachments: YARN-4101.1.patch, YARN-4101.2.patch
> Currently, There is no way for user to understand Zk-RM has connection 
> issues. In HA environment, RM is highly dependent on Zookeeper. If connection 
> between RM and Zk is jeopardized, cluster is likely to be gone in bad state.
> Example: Rm1 is active and Rm2 is standby. If connection between Rm2 and Zk 
> is lost, Rm2 will never become active. In this case, if Rm1 hits an error and 
> could not be started, cluster goes in bad state. This situation is very hard 
> to debug for user. In this case, if we can develop better prompting of 
> messages, User could fix the Zk-RM connection issue and could avoid getting 
> in bad state.
> Thus, We need a better way to prompt alert to user if connection between Zk 
> -> Active RM or Zk -> standby RM is getting bad.
> Here are the suggestions.
> 1) Print connection lost alert in RM UI
> 2) Print alert messages while running any Yarn command such as yarn logs, 
> yarn applications etc

This message was sent by Atlassian JIRA

Reply via email to