Masatake Iwasaki commented on YARN-3805:

in NodeManager.md:
once a disk is marked as unhealthy, the NodeManager stops checking it to see if 
it has recovered(e.g. disk became full and was then cleaned up). The only way 
for the NodeManager to use that disk to restart the software on the node.

> Update the documentation of Disk Checker based on YARN-90
> ---------------------------------------------------------
>                 Key: YARN-3805
>                 URL: https://issues.apache.org/jira/browse/YARN-3805
>             Project: Hadoop YARN
>          Issue Type: Bug
>          Components: documentation
>            Reporter: Masatake Iwasaki
>            Assignee: Masatake Iwasaki
>            Priority: Minor
> NodeManager is able to recover status of the disk once broken and fixed 
> without restarting by YARN-90.

This message was sent by Atlassian JIRA

Reply via email to