[HDFS-15001] Automatically add back the failed volume if it gets better again

• Type: Improvement
• Status: Patch Available
• Priority: Major
• Resolution: Unresolved
• Component/s: datanode
• Assignee: Yang Yun
• Reporter: Yang Yun
• Created: 21/Nov/19 08:44
• Updated: 21/Nov/19 11:38

Description

For any reason, if one volume is regard as failed and removed from datanode, we have to restart the datanode to add it back.
I add  a more check on background for failed volume, and add them back automatically if it gets better again.
Reuse the refreshVolumes logical for the old way can not delete all things related the failed valume. 
Add a new method checkFailedVolume to DatasetVolumeChecker to schedule task for checking failed volume repeatly.