[HBASE-1086] If sidekick datanode goes down, regionserver is hosed

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Invalid

Environment

streamy cluster, recent TRUNK

• Assignee: Unassigned
• Reporter: Michael Stack
• Created: 23/Dec/08 18:35
• Updated: 11/Jun/22 20:30
• Resolved: 01/Oct/10 23:32

Description

On the jgray cluster, datanodes died. Subsequently, could read from the table fine but attempts at writing into regions hosted by regionservers that sat beside dead datanodes fail – connection refused.

This issue should be easy to replicate, just kill adjacent datanode.