[HDFS-8699] DFSClient hang up when there are not sufficient DataNodes in EC cluster.

• Type: Sub-task
• Status: Resolved
• Priority: Major
• Resolution: Duplicate
• Affects Version/s: HDFS-7285
• Target Version/s: HDFS-7285
• Assignee: Kai Sasaki Assign to me
• Reporter: Kai Sasaki
• Created: 30/Jun/15 08:02
• Updated: 30/Jun/15 09:33
• Resolved: 30/Jun/15 09:33

Description

6-3 RS schema requires 9 data nodes at least to work properly. However when there are not sufficient DataNodes in EC cluster, writing file fails because of NullPointerException. And also client is hang up.

java.lang.NullPointerException
        at org.apache.hadoop.hdfs.DataStreamer.nextBlockOutputStream(DataStreamer.java:1463)
        at org.apache.hadoop.hdfs.DataStreamer.run(DataStreamer.java:591)
        at org.apache.hadoop.hdfs.StripedDataStreamer.run(StripedDataStreamer.java:49)

DFSClient should handle insufficient DN error properly just like replication blocks.

Issue Links

• duplicates: HDFS-8220 Erasure Coding: StripedDataStreamer fails to handle the blocklocations which doesn't satisfy BlockGroupSize