[HDFS-4135] Add method to DFSOutputStream to dump pipeline status

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 2.0.2-alpha, 3.0.0-alpha1
• Component/s: hdfs-client
• Assignee: Unassigned
• Reporter: Todd Lipcon
• Created: 01/Nov/12 17:06
• Updated: 12/May/16 18:13

Description

Occasionally higher-level apps (like HBase) will detect that an outputstream is acting slower than usual. It would be nice to be able to log a WARN message which contains the current state of the pipeline, so that its easier for operators to diagnose why things might be acting up. I would like to see:

• block ID
• number of acked bytes, number of bytes written to pipeline
• current state of ack queue (how many packets are un-acked)
• datanode IDs of pipeline
• any latency information we could easily collect