[HIVE-16478] The result of select sum sql is not correct when some datanodes of hadoop cluster are sometime bad and sometime good every 1-2mins

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 1.1.0

Environment

Linux 2.6.32-431.el6.x86_64, CDH Cluster

• Assignee: Hive QA
• Reporter: zhao hongjun
• Created: 19/Apr/17 09:32
• Updated: 19/Apr/17 09:52

Description

1. submiting a select sum sql to hadoop cluster with hive

2. the mapreduce job are running on the datanodes: DN_A, DN_B, DN_C

3. When map job is completed, the datanodes DN_A, DN_B, DN_C are bad health(disk utilization percent reach the value 'yarn.nodemanager.disk-health-checker.max-disk-utilization-per-disk-percentage' ), and reported to resource manager. At the same time, application master killed the map job.

4. the result of select sum sql is not correct.