[HADOOP-14348] Need HDFS Health scripts for monitoring datanode

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: scripts
• Assignee: Unassigned
• Reporter: Rahul Shrivastava
• Created: 24/Apr/17 06:10
• Updated: 24/Apr/17 06:11

Description

Sick machines cause issues, that HDFS will not automatically remediate that.
Upon a bad health, the DN should either stop sending block reports, simply shut itself down, or indicate the problem in some other way.
We need a script to achieve this for DN.

MapReduceV1 had something, see https://issues.apache.org/jira/browse/MAPREDUCE-211 . Hadoop-common moved into hadoop-yarn-project for Hadoop v2, meaning we have a solution for YARN but are lacking what we need in HDFS.