[jira] [Created] (REEF-865) Make FailTask-related testsuite robust by adding sleeping to wait for tasks ready

["Dongjoon Hyun (JIRA)" <ji...@apache.org>]

Dongjoon Hyun created REEF-865:
----------------------------------

             Summary: Make FailTask-related testsuite robust by adding sleeping to wait for tasks ready 
                 Key: REEF-865
                 URL: https://issues.apache.org/jira/browse/REEF-865
             Project: REEF
          Issue Type: Improvement
          Components: REEF-Tests
    Affects Versions: 0.14
            Reporter: Dongjoon Hyun
            Assignee: Dongjoon Hyun
            Priority: Minor


FailTaskStop/FailTaskClose fails frequently in Docker environment with the following error messages.
 
{code}
org.apache.reef.runtime.common.driver.defaults.DefaultClientCloseHandler onNext
WARNING: Received a close message from the client, but no handler was bound for it. Active threads:
{code}

In a successful case, FailTaskStop/FailTaskClose register their handlers. In a error case like the above, 'registering' is not executed yet. So, `DefaultClientCloseHandler.onNext()` is invoked and the test case hangs until its timeout occurs.

By adding a sleep time in `Driver.java` of this testsuite, we can make the above test cases robust on resource-limited environments. 
{code}
      // In case of slow machines, it takes some time until task is ready.
      try {
        Thread.sleep(1000);
      } catch (final InterruptedException ex) {
        LOG.log(Level.INFO, "Interrupted: {0}", ex);
      }
{code}

Note, this does not change any functionality of this testcases.

I experimented this fix on my labtop docker and travis-ci.org 10 times. 

https://travis-ci.org/dongjoon-hyun/incubator-reef/builds



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)