[RATIS-1203] Fix flaky tests caused by retryCache in TestRetryCacheMetrics.java

• Type: Improvement
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 2.0.0
• Component/s: server
• Assignee: Unassigned
• Reporter: KellyShao
• Created: 04/Dec/20 04:32
• Updated: 04/Dec/20 05:49
• Resolved: 04/Dec/20 05:49

Description

• Issue:
  • The method testRetryCacheEntryCount in TestRetryCacheMetrics.java class is flaky. Everytime when we trigger testRetryCacheHitMissCount method before testRetryCacheEntryCount, the testRetryCacheEntryCount will failed with error message: 

    [ERROR] test1RetryCacheEntryCount(org.apache.ratis.server.impl.TestRetryCacheMetrics)  Time elapsed: 0.004 s  <<< FAILURE!
    java.lang.AssertionError: expected:<1> but was:<0>
    

• Root cause:
  • The root cause is that the variable retryCache in this class is only initialized once in @BeforeClass, and cleared only at the end of the testRetryCacheEntryCount method. But, the testRetryCacheHitMissCount doesn't clear it.
  • So, when testRetryCacheHitMissCount is triggered before testRetryCacheEntryCount, the line checkEntryCount(0); within testRetryCacheEntryCount will failed.

• Solution:
  • Added an @After method in order to tear down retryCache after each test.

    @After
    public void tearDown() {
        retryCache.close();
        checkEntryCount(0);
    }      

  • Deleted 2 lines of code in testRetryCacheEntryCount to avoid duplication.

    retryCache.close();
    checkEntryCount(0);
    

• The flaky test is found by running https://github.com/idflakies/iDFlakies

Issue Links

• links to: GitHub Pull Request #321