> On March 6, 2014, 7:05 p.m., Suman Karumuri wrote:
> > If I remember the discussion correctly, there is a race condition here, 
> > where some actions could still be performed through the memory store, 
> > between the time the storage is shutdown and the shutdown actually happens. 
> > That prompted us to do this change at the storage layer instead. Is there a 
> > reason, we are doing this change at the log layer now?
> >

I can't recall the specifics that led us there, but the biggest issue i can see 
is that nested transactions don't honor the happens-before expectation between 
storage write and side-effects.  Given that we already face this issue, i'm 
leaning towards keeping this behavior where it is in the diff, and addressing 
that bug separately.


- Bill


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/18714/#review36401
-----------------------------------------------------------


On March 3, 2014, 11:35 p.m., Bill Farner wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/18714/
> -----------------------------------------------------------
> 
> (Updated March 3, 2014, 11:35 p.m.)
> 
> 
> Review request for Aurora, Kevin Sweeney and Suman Karumuri.
> 
> 
> Bugs: AURORA-221
>     https://issues.apache.org/jira/browse/AURORA-221
> 
> 
> Repository: aurora
> 
> 
> Description
> -------
> 
> Disable log and initiate shutdown upon log storage failures.
> 
> 
> Diffs
> -----
> 
>   src/main/java/org/apache/aurora/scheduler/log/mesos/MesosLog.java 
> 071c4fcf96099819aac6276306034434e3eec879 
>   src/test/java/org/apache/aurora/scheduler/log/mesos/MesosLogTest.java 
> 359f5d44e030a53cedd612a350b161b392d929c4 
> 
> Diff: https://reviews.apache.org/r/18714/diff/
> 
> 
> Testing
> -------
> 
> ./gradlew build
> 
> 
> Thanks,
> 
> Bill Farner
> 
>

Reply via email to