[HADOOP-18633] fix test AbstractContractDistCpTest#testDistCpUpdateCheckFileSkip

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 3.3.6
• Component/s: tools/distcp
• Labels: pull-request-available
• Target Version/s: 3.3.9
• Assignee: Mehakmeet Singh
• Reporter: Mehakmeet Singh
• Created: 15/Feb/23 09:50
• Updated: 08/Jun/23 20:25
• Resolved: 22/Feb/23 09:02

Description

In the newly introduced test testDistCpUpdateCheckFileSkip, for the first pass of "distcp -update", target file should not be present so that the copy takes place and creates the target file. 

Currently, we create both the source and target file with same block size from the start which can lead to flakiness due to race condition causing the modification time of the target file to be greater than/equal to the source and not copy the file at all. This can be seen more in the TestLocalContractDistCp due to no remote calls to create the target.

java.lang.AssertionError: Mismatch in COPY counter value expected:<1> but was:<0>
	at org.junit.Assert.fail(Assert.java:89)
	at org.junit.Assert.failNotEquals(Assert.java:835)
	at org.junit.Assert.assertEquals(Assert.java:647)
	at org.apache.hadoop.tools.contract.AbstractContractDistCpTest.verifySkipAndCopyCounter(AbstractContractDistCpTest.java:1000)
	at org.apache.hadoop.tools.contract.AbstractContractDistCpTest.testDistCpUpdateCheckFileSkip(AbstractContractDistCpTest.java:919) 

Issue Links

• is caused by: HADOOP-18596 Distcp -update between different cloud stores to use modification time while checking for file skip.
• links to: GitHub Pull Request #5401
• links to: GitHub Pull Request #5422