> On Feb. 16, 2016, 11:02 a.m., Alexander Rukletsov wrote:
> > Could you please explain in description why disabling checkpointing speeds 
> > up the test and why it is ok to disable it for these tests?
> 
> haosdent huang wrote:
>     Sure!
> 
> Bernd Mathiske wrote:
>     Not writing to disk costs less time.
>     
>     If a test does not attempt recovery, then whatever would be checkpointed 
> never gets used.
>     
>     Only one test attempts/tests recovery so far. In this test we do apply 
> checkpointing. So it does get tested, but we don't need to test the writing 
> part of it in every other test.
> 
> Alexander Rukletsov wrote:
>     I'm a bit confused about what contributes most here. Is it checkpointing 
> or rather the other change? It's a bit strange to me that checkpointing 
> significantly influences the test duration. If so, I would like to understand 
> why, of not, them maybe we can pull it out of the change to avoid 
> distraction. Also It's not clear at the first glance why it's ok to reduce 
> some loops in the tests and why they were there in the first place.
>     
>     it would be great to have one fix per patch with explanation why it helps 
> and how much.

>it would be great to have one fix per patch with explanation why it helps and 
>how much.

Sure. My test machine is use SATA disk. And becsue we write to disk when 
status_update, would cause >100 ms delay when I testing. I agree with you. Let 
me split up them and explain why it helps and how much.


- haosdent


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/42696/#review119294
-----------------------------------------------------------


On Feb. 16, 2016, 11:39 a.m., haosdent huang wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/42696/
> -----------------------------------------------------------
> 
> (Updated Feb. 16, 2016, 11:39 a.m.)
> 
> 
> Review request for mesos and Bernd Mathiske.
> 
> 
> Bugs: MESOS-4486
>     https://issues.apache.org/jira/browse/MESOS-4486
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Speed up FetcherCacheTest.Local* test cases.
> 
> 
> Diffs
> -----
> 
>   src/tests/fetcher_cache_tests.cpp 1cf45660691860793ac600363f7934e13a2e7ddf 
> 
> Diff: https://reviews.apache.org/r/42696/diff/
> 
> 
> Testing
> -------
> 
> Before apply the patch:
> ```
> $ sudo ./bin/mesos-tests.sh --gtest_filter="FetcherCacheTest.Local*"
> [       OK ] FetcherCacheTest.LocalUncached (2580 ms)
> [       OK ] FetcherCacheTest.LocalCached (2524 ms)
> [       OK ] FetcherCacheTest.LocalUncachedExtract (2514 ms)
> [       OK ] FetcherCacheTest.LocalCachedExtract (2551 ms)
> ```
> 
> 
> After apply the patch:
> ```
> $ sudo ./bin/mesos-tests.sh --gtest_filter="FetcherCacheTest.Local*"
> [       OK ] FetcherCacheTest.LocalUncached (873 ms)
> [       OK ] FetcherCacheTest.LocalCached (1609 ms)
> [       OK ] FetcherCacheTest.LocalUncachedExtract (926 ms)
> [       OK ] FetcherCacheTest.LocalCachedExtract (1509 ms)
> ```
> 
> Also test the recovery test manually:
> 
> ```
> sudo ./bin/mesos-tests.sh 
> --gtest_filter="FetcherCacheHttpRecoveryTest.HttpCachedRecovery"
> ```
> 
> 
> Thanks,
> 
> haosdent huang
> 
>

Reply via email to