[HDFS-8517] Fix a decoding issue in stripped block recovering in client side

• Type: Sub-task
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: HDFS-7285
• Hadoop Flags: Reviewed
• Assignee: Kai Zheng
• Reporter: Kai Zheng
• Created: 02/Jun/15 13:14
• Updated: 30/Sep/15 19:38
• Resolved: 02/Jun/15 22:37

Description

jingzhao reported a decoding issue in HDFS-8481 in the comment copied below:

While debugging HDFS-8319, I just found that in TestWriteReadStripedFile#testWritePreadWithDNFailure, if we change the startOffsetInFile from cellSize * 5 to 0, the test fails with the following error msg:

java.lang.AssertionError: Byte at 524288 should be the same expected:<27> but was:<-9>
	at org.junit.Assert.fail(Assert.java:88)
	at org.junit.Assert.failNotEquals(Assert.java:743)
	at org.junit.Assert.assertEquals(Assert.java:118)
	at org.junit.Assert.assertEquals(Assert.java:555)
	at org.apache.hadoop.hdfs.TestWriteReadStripedFile.testWritePreadWithDNFailure(TestWriteReadStripedFile.java:390)

It was caused by an issue that we need to adjust the inputs order to call decoder#decode. When startOffsetInFile is 5 * cellSize, the recovering or decoding won't be needed as ZERO cells don't need to recover; when it's 0, the issue then was exposed, as normal data cells must be recovered thru the decoding.