aurora-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Kevin Sweeney" <kevi...@apache.org>
Subject Re: Review Request 16636: Ensure SchedulerActive event is dispatched before leader is advertised.
Date Mon, 06 Jan 2014 19:30:19 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/16636/#review31246
-----------------------------------------------------------

Ship it!


Ship It!

- Kevin Sweeney


On Jan. 4, 2014, 3:21 p.m., Bill Farner wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/16636/
> -----------------------------------------------------------
> 
> (Updated Jan. 4, 2014, 3:21 p.m.)
> 
> 
> Review request for Aurora, Kevin Sweeney and Maxim Khutornenko.
> 
> 
> Repository: aurora
> 
> 
> Description
> -------
> 
> I dug into this after noticing SchedulerIT test flakiness (deadlock) as a result of tearDown
actions being executed SchedulerActive event was being dispatched.  The test is arranged such
that it expects the effects of that event to occur before the test exits.  The deadlock was
related to EasyMock's default global lock for mock answering, so that particular deadlock
does not affect us in production.
> 
> Please see the lengthy comment in SchedulerLifecycle.java explaining the mechanics of
the change.
> 
> 
> Diffs
> -----
> 
>   src/main/java/org/apache/aurora/scheduler/SchedulerLifecycle.java bf926a740c3dafc811165e13a437853a5390bfe6

>   src/test/java/org/apache/aurora/scheduler/SchedulerLifecycleTest.java 24cd6faa4c9610b369002e842450b9b2bf87dfca

>   src/test/java/org/apache/aurora/scheduler/app/SchedulerIT.java 41fd3a0672a264a93814b1e2d1585232d9d506f9

> 
> Diff: https://reviews.apache.org/r/16636/diff/
> 
> 
> Testing
> -------
> 
> Prior to this change, i could trigger the deadlock about 50% of the time.  With this
change, i've run tests many times with no issues.
> 
> 
> Thanks,
> 
> Bill Farner
> 
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message