mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ben Mahler" <benjamin.mah...@gmail.com>
Subject Re: Review Request 39873: Fixed master to properly handle status updates when multiple of them are enqueued on the slave.
Date Wed, 04 Nov 2015 02:43:52 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/39873/#review105027
-----------------------------------------------------------



src/master/master.cpp (lines 6004 - 6008)
<https://reviews.apache.org/r/39873/#comment163394>

    Just looking at the patch that introduced this bug, why are we removing the out-of-order
update prevention? Don't see any mention of this in the stuff around https://issues.apache.org/jira/browse/MESOS-2864.



src/master/master.cpp (line 6034)
<https://reviews.apache.org/r/39873/#comment163395>

    Why not have the same logic here to be defensive? Or do you intend to guard against it?
Just seems weird to allow it in one of the cases but not the other.


- Ben Mahler


On Nov. 2, 2015, 10:34 p.m., Vinod Kone wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/39873/
> -----------------------------------------------------------
> 
> (Updated Nov. 2, 2015, 10:34 p.m.)
> 
> 
> Review request for mesos and Ben Mahler.
> 
> 
> Bugs: MESOS-2864
>     https://issues.apache.org/jira/browse/MESOS-2864
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> The master now doesn't change the latest state of a task if it has already terminated.
But it still updates the status update state and uuid.
> 
> 
> Diffs
> -----
> 
>   src/master/master.cpp 2bc5a97a5b50c8a8a9902c47b2e9e3b5216d97ea 
> 
> Diff: https://reviews.apache.org/r/39873/diff/
> 
> 
> Testing
> -------
> 
> make check
> 
> Ran the "RecoverCompletedExecutor" test 100 times as it was failing most of the time
without this fix.
> 
> 
> Thanks,
> 
> Vinod Kone
> 
>


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