[HADOOP-16677] Recalculate the remaining timeout millis correctly while throwing an InterupptedException in SocketIOWithTimeout.

• Type: Bug
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 3.1.3
• Fix Version/s: 3.3.0, 3.1.4, 3.2.2
• Component/s: common
• Labels: pull-request-available
• Assignee: Xudong Cao
• Reporter: Xudong Cao
• Created: 31/Oct/19 07:48
• Updated: 12/Nov/19 22:34
• Resolved: 12/Nov/19 22:28

Description

In SocketIOWithTimeout, when a thread was interrupted and exit from select(), it proceed to throw an InterruptedIOException, in exception message the remaining timeout mills should be calcuated correctly rather than simply give a total timeout millis ,  which could be very misleading.

 

An example log before this jira:

2019-10-31 16:20:39,496 INFO org.apache.hadoop.hdfs.server.datanode.DataNode: Exception for BP-1895911599-192.168.202.11-1572488735889:blk_1073741833_1016
java.io.InterruptedIOException: Interrupted while waiting for IO on channel java.nio.channels.SocketChannel[connected local=/192.168.202.12:50010 remote=/192.168.202.11:57006]. 60000 millis timeout left.
at org.apache.hadoop.net.SocketIOWithTimeout$SelectorPool.select(SocketIOWithTimeout.java:351)
at org.apache.hadoop.net.SocketIOWithTimeout.doIO(SocketIOWithTimeout.java:157))

 

An example log after this jira:

2019-10-31 16:20:39,496 INFO org.apache.hadoop.hdfs.server.datanode.DataNode: Exception for BP-1895911599-192.168.202.11-1572488735889:blk_1073741833_1016
java.io.InterruptedIOException: Interrupted while waiting for IO on channel java.nio.channels.SocketChannel[connected local=/192.168.202.12:50010 remote=/192.168.202.11:57006]. Total timeout mills is 60000, 1000 millis timeout left.
at org.apache.hadoop.net.SocketIOWithTimeout$SelectorPool.select(SocketIOWithTimeout.java:351)
at org.apache.hadoop.net.SocketIOWithTimeout.doIO(SocketIOWithTimeout.java:157))

Issue Links

• links to: GitHub Pull Request #1687