[HDFS-656] Clarify error handling and retry semantics for DFS read path

• Type: Task
• Status: Open
• Priority: Minor
• Resolution: Unresolved
• Component/s: hdfs-client
• Assignee: Unassigned
• Reporter: Todd Lipcon
• Created: 25/Sep/09 05:23
• Updated: 25/Sep/09 06:02

Description

HDFS-127 has brought to light a lot of confusion regarding the DFS read semantics. This JIRA is to define what we think the semantics should be, and potentially also to discover what they are now. It would be good to produce a document so we are later able to fix or reimplement the DFS read path and avoid confusion in the future.

Issue Links

• is related to: HDFS-50 NullPointerException when reading deleted file
• is related to: HDFS-127 DFSClient block read failures cause open DFSInputStream to become unusable
• is related to: HDFS-378 DFSClient should track failures by datanode across all streams
• is related to: HDFS-575 DFSClient read performance can be improved by stagerring connection setup to datanode(s)
• is related to: HDFS-380 support for persistent connections to improve random read performance.