> On Jan. 31, 2014, 2:55 a.m., Kevin Sweeney wrote:
> > src/main/java/org/apache/aurora/scheduler/async/TaskScheduler.java, lines 
> > 185-187
> > <https://reviews.apache.org/r/17578/diff/2/?file=457399#file457399line185>
> >
> >     I think this is actually somewhat a regression - in the past it was 
> > possible for a "cheap" filter to reject a task just based on a precomputed 
> > index and avoid a query altogether. Now we will always make at least one 
> > query. How would you feel about making this query lazy so that it happens 
> > the first time the info is needed rather than eagerly when it might not be 
> > needed?

In the spirit of the interfaces, you're correct.  In practice, though, it's 
going to be exceedingly rare that we don't do this query at least once in a 
scheduling round.  It would only happen if there are no resource offers, or if 
all offer/task combinations have dedicated mismatches.

Thinking forward, though, laziness will avoid unexpected behavior, so i've 
added it.


- Bill


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/17578/#review33285
-----------------------------------------------------------


On Jan. 31, 2014, 2:17 a.m., Bill Farner wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/17578/
> -----------------------------------------------------------
> 
> (Updated Jan. 31, 2014, 2:17 a.m.)
> 
> 
> Review request for Aurora, Kevin Sweeney and Maxim Khutornenko.
> 
> 
> Bugs: AURORA-139
>     https://issues.apache.org/jira/browse/AURORA-139
> 
> 
> Repository: aurora
> 
> 
> Description
> -------
> 
> The majority of the code here is related to plumbing the argument through two 
> layers (SchedulingFilter, TaskAssigner) and updating related tests.
> 
> One functional change in this diff is the removal of a filter in 
> AttributeFilter, which was unnecessary since the query was already 
> job-scoped.  The filter existed because at one point we supported job 
> configurations that could request to avoid other jobs, so the task set would 
> include other jobs.
> 
> 
> Diffs
> -----
> 
>   src/main/java/org/apache/aurora/scheduler/async/AsyncModule.java 
> 72d3621055d28b7b0ec7eeca7a1e8be6ce1e5042 
>   src/main/java/org/apache/aurora/scheduler/async/Preemptor.java 
> c11f4837152c3c8de07058d35a874b406ea7878f 
>   src/main/java/org/apache/aurora/scheduler/async/TaskScheduler.java 
> 4afc33259cc0150424f8d790a38c0d0b38fb2392 
>   
> src/main/java/org/apache/aurora/scheduler/events/NotifyingSchedulingFilter.java
>  c7f4a1b975129c58328e42c758683ebad06ae035 
>   src/main/java/org/apache/aurora/scheduler/filter/AttributeFilter.java 
> 0816d3f99ce8347d6ef705116b48abb164854c8a 
>   src/main/java/org/apache/aurora/scheduler/filter/CachedJobState.java 
> PRE-CREATION 
>   src/main/java/org/apache/aurora/scheduler/filter/ConstraintFilter.java 
> 17fc8b99c48b3133af3d1f0c3c75d822e78757e6 
>   src/main/java/org/apache/aurora/scheduler/filter/SchedulingFilter.java 
> 999e0f7bba882a52a6b278f36c26c48a87a88958 
>   src/main/java/org/apache/aurora/scheduler/filter/SchedulingFilterImpl.java 
> 4ba1483386e8fe988a800253c44e9359561bd972 
>   src/main/java/org/apache/aurora/scheduler/state/TaskAssigner.java 
> d8f326e45aaacbff3447c3d5b12fb4f9112a82cf 
>   src/test/java/org/apache/aurora/scheduler/async/PreemptorImplTest.java 
> 025294fd51abe1eb41a6ade39b5a9fea8a8422b6 
>   src/test/java/org/apache/aurora/scheduler/async/TaskSchedulerImplTest.java 
> 35dd666c44f39d6b13e4dc516fed8117f88db0e5 
>   src/test/java/org/apache/aurora/scheduler/async/TaskSchedulerTest.java 
> 6c124c53130530881699abfd5634ba0171932afa 
>   
> src/test/java/org/apache/aurora/scheduler/events/NotifyingSchedulingFilterTest.java
>  c5379fe725a5dd36dbe6217c18b281d7a3209c67 
>   
> src/test/java/org/apache/aurora/scheduler/filter/SchedulingFilterImplTest.java
>  f3df73c1d248165bf19bdfd1305fd7a9b8190cf7 
>   src/test/java/org/apache/aurora/scheduler/state/TaskAssignerImplTest.java 
> 154ccc4716168ad966b246280370600de92d5d7f 
> 
> Diff: https://reviews.apache.org/r/17578/diff/
> 
> 
> Testing
> -------
> 
> ./gradlew build
> 
> 
> Thanks,
> 
> Bill Farner
> 
>

Reply via email to