Peter Bacsko commented on OOZIE-2585:

[~gezapeti] I rebased the patch. As we discussed IRL, I removed the two tests 

> Disable or remove test case 
> TestPartitionDependencyManagerEhcache.testMemoryUsageAndSpeedOverflowToDisk 
> and testMemoryUsageAndSpeed
> -----------------------------------------------------------------------------------------------------------------------------------
>                 Key: OOZIE-2585
>                 URL: https://issues.apache.org/jira/browse/OOZIE-2585
>             Project: Oozie
>          Issue Type: Sub-task
>            Reporter: Peter Bacsko
>            Assignee: Peter Bacsko
>            Priority: Minor
>         Attachments: OOZIE-2585-001.patch, OOZIE-2585-002.patch, 
> OOZIE-2585-003.patch
> There is a problem with the test case testMemoryUsageAndSpeedOverflowToDisk.
> After 20k entries, data is spilled to the disk. However, Ehcache writes on a 
> separate thread asynchronously and all put operations return immediately. The 
> problem is that after storing the 60k entries, we try to read them back 
> instantly, with many of them are still being flushed to the storage. As long 
> as disk write in progress, these entries are unavailable.
> I didn't find a reliable way to wait for disk writes. Therefore I suggest 
> disabling this test or eliminate it altogether -- it's a performance test 
> that runs on all kinds of machines during a build, I don't think it makes too 
> much sense to keep it.
> Also testMemoryUsageAndSpeed might be prone to test failures too - these two 
> assertions might evaluate to false:
> {code}
> assertTrue((endTime - startTime) < insertTimeinMillis);
> ...
> assertTrue((System.currentTimeMillis() - endTime) < retrievalTimeinMillis);
> {code}

This message was sent by Atlassian JIRA

Reply via email to