> On May 19, 2014, 9:11 p.m., Mark Chu-Carroll wrote:
> > src/main/java/org/apache/aurora/scheduler/state/SchedulerCoreImpl.java, 
> > line 147
> > <https://reviews.apache.org/r/21383/diff/3/?file=584300#file584300line147>
> >
> >     Just to be clear: there is no deprecation here. The existing behavior 
> > will continue to work. We will deprecate the old cron behavior eventually, 
> > but not for a while yet.
> >     
> >     Also: is there any documented version of this procedure? I haven't seen 
> > or heard of it before, so I assume others haven't either?
> >
> 
> Bill Farner wrote:
>     Is there any reason to not deprecate now?
>     
>     We _literally_ made it up on Friday afternoon, so no - no docs yet.  We 
> haven't even discussed it with anyone, so i think we're in the 'see how it 
> works' phase.
> 
> Mark Chu-Carroll wrote:
>     Yes, there's a reason not to deprecate now: I don't want this to be a 
> sudden change that disrupts all of our aurora users. I'd like to start by 
> making a new alternative way of scheduling cron jobs, tell people, and give 
> them some deprecation period during which to transition to the new method, 
> before we just pull the old way out from under them.
>     
>     This change is going to have a big impact on users. I think in the long 
> run it'll be positive, but it's the kind of thing where if it's not done 
> gracefully, people will justifiably be very upset.

I think we want the same thing :-)

My proposal is not to remove the support, but to start the deprecation cycle 
now, and remove the feature in 0.7.0.


> On May 19, 2014, 9:11 p.m., Mark Chu-Carroll wrote:
> > src/main/java/org/apache/aurora/scheduler/thrift/SchedulerThriftInterface.java,
> >  line 258
> > <https://reviews.apache.org/r/21383/diff/3/?file=584301#file584301line258>
> >
> >     There are tests of scheduleCronJob and descheduleCronJob below. Did you 
> > not notice those, or is there a particular kind of test that you're looking 
> > for?
> >
> 
> Bill Farner wrote:
>     Those tests cover SchedulerCoreImpl, but not SchedulerThriftInterface.  
> We'll need to cover the code in both.
> 
> Maxim Khutornenko wrote:
>     Bill Farner, any reason we should continue investing into SchedulerCore? 
> I was under assumption we were trying to get rid of it.

Thanks, that's a good point.  We've been trying to shy away from cases where 
SchedulerCoreImpl is just a proxy to another class.  Maxim correctly points out 
that SchedulerCoreImpl is just a front for CronJobManager, with little 
additional logic.  It's probably best to just push this code back out to 
SchedulerThriftInterface.


- Bill


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/21383/#review43397
-----------------------------------------------------------


On May 19, 2014, 3:25 p.m., Mark Chu-Carroll wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/21383/
> -----------------------------------------------------------
> 
> (Updated May 19, 2014, 3:25 p.m.)
> 
> 
> Review request for Aurora, David McLaughlin and Bill Farner.
> 
> 
> Bugs: aurora-417
>     https://issues.apache.org/jira/browse/aurora-417
> 
> 
> Repository: aurora
> 
> 
> Description
> -------
> 
> Add cron schedule and deschedule calls to the scheduler API.
> 
> 
> Diffs
> -----
> 
>   src/main/java/org/apache/aurora/scheduler/state/SchedulerCore.java 
> 226b71ce7749492abd3e1d673382668c9011a1c8 
>   src/main/java/org/apache/aurora/scheduler/state/SchedulerCoreImpl.java 
> d377974da4f7efa7f47e06702654bc786299e01c 
>   
> src/main/java/org/apache/aurora/scheduler/thrift/SchedulerThriftInterface.java
>  4386a86c0aebdb9666c0232672f8c1bab7458f47 
>   src/main/thrift/org/apache/aurora/gen/api.thrift 
> 66292dc0369941fc62719d49209edc07adc81a53 
>   
> src/test/java/org/apache/aurora/scheduler/state/BaseSchedulerCoreImplTest.java
>  0e17f49e68953b5f043301dbb1b46b96968e1247 
>   src/test/java/org/apache/aurora/scheduler/thrift/aop/ForwardingThrift.java 
> 933a56bf3f7165fa84aedcc8d1392e32824fd487 
>   src/test/python/apache/aurora/client/api/test_scheduler_client.py 
> 399d4c614dda93ae0e7d8f38a02379bc0f42e4af 
> 
> Diff: https://reviews.apache.org/r/21383/diff/
> 
> 
> Testing
> -------
> 
> Ran all tests with gradlew; no errors. (Includes two new tests, that ensure 
> that calling the new APIs causes the right internal cron methods to be 
> called.)
> 
> 
> Thanks,
> 
> Mark Chu-Carroll
> 
>

Reply via email to