Re: TestSlabCache

[Li Pi <lp...@ucsd.edu>]

posting it now.

On Sun, Sep 18, 2011 at 5:03 PM, Ted Yu <yu...@gmail.com> wrote:
> Please reply to dev@ today.
> Stack is waiting to branch 0.92
>
> Thanks
>
>
>
> On Sep 18, 2011, at 2:07 PM, Li Pi <lp...@ucsd.edu> wrote:
>
>> Yeah, I'll add that right now.
>>
>> On Sun, Sep 18, 2011 at 2:05 PM, Ted Yu <yu...@gmail.com> wrote:
>>> I would use the following instead of System.print():
>>>
>>>   public static final Log LOG =
>>>     LogFactory.getLog(CacheTestUtils.class);
>>>
>>> I think build log would be flooded by TestSlabCache.
>>>
>>> Shall we introduce a separate config parameter for enabling off-heap cache ?
>>> Currently we have:
>>>     long offHeapCacheSize = (long)
>>> (conf.getFloat("hbase.offheapcache.percentage", (float) 0.95) *
>>> DirectMemoryUtils.getDirectMemorySize());
>>> Somebody would say what if DirectMemorySize is specified but user wants to
>>> turn off off-heap cache ?
>>>
>>> Thanks
>>>
>>> On Sun, Sep 18, 2011 at 1:54 PM, Li Pi <lp...@ucsd.edu> wrote:
>>>>
>>>> If I put a printout - will it show up in the build logs?
>>>>
>>>> On Sun, Sep 18, 2011 at 1:17 PM, Ted Yu <yu...@gmail.com> wrote:
>>>>> Li:
>>>>> Can CacheTestUtils.hammerSingleKey() and CacheTestUtils.hammerEviction()
>>>>> be
>>>>> improved w.r.t. logging ?
>>>>> Since we don't have access to Jenkins build machine and the test is
>>>>> hanging,
>>>>> logging seems to be the only way.
>>>>>
>>>>> Thanks
>>>>>
>>>>> On Sun, Sep 18, 2011 at 1:02 PM, Li Pi <lp...@ucsd.edu> wrote:
>>>>>>
>>>>>> I'm unable to reproduce this somehow - not sure why. The test
>>>>>> completes if tested alone - but not when all tests are run? I'll try
>>>>>> to jstack it during that particular test.
>>>>>>
>>>>>>
>>>>>> On Sun, Sep 18, 2011 at 12:26 PM, Andrew Purtell <ap...@apache.org>
>>>>>> wrote:
>>>>>>>> testCacheMultiThreadedEviction() could be the real issue. If we
>>>>>>>> don't
>>>>>>>> have
>>>>>>>> time to experiment further, we can just disable TestSlabCache.
>>>>>>>
>>>>>
>>>
>>>
>