> On 八月 23, 2016, 8:45 a.m., Guangya Liu wrote:
> > src/master/master.cpp, line 3366
> > <https://reviews.apache.org/r/51317/diff/1/?file=1481628#file1481628line3366>
> >
> > I saw that this only occur if the task authorization failed for now, shall we
highlight this in the comments here?
>
> Benjamin Mahler wrote:
> Hm.. it can occur if authorization fails or if validation fails, why did you mention
only authorization failure?
>
> We could update the code to avoid sending TASK_ERROR REASON_UNAUTHORIZED when the
task is not in the `pendingTasks` map, since we know that it is either killed or invalid,
but this seemed not worth it since it only half-way fixes the issue.
>
> Guangya Liu wrote:
> Yes, but in `_accept`, we only send out `TASK_ERROR` when authorization failed but
did not handle the case if `validation failed`, what about put `validation failed` in `TODO`?
We already have a `TODO` in `_accept`, I think it is good enough to handle the question here.
Thanks.
- Guangya
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/51317/#review146492
-----------------------------------------------------------
On 八月 23, 2016, 5:11 a.m., Benjamin Mahler wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/51317/
> -----------------------------------------------------------
>
> (Updated 八月 23, 2016, 5:11 a.m.)
>
>
> Review request for mesos and Vinod Kone.
>
>
> Repository: mesos
>
>
> Description
> -------
>
> The use of `pendingTasks` cannot distinguish between a duplicate
> TaskID and a task that has been killed while pending. This means
> that if an invalid or unauthorized task is killed while pending,
> TASK_KILLED is sent, and once in Master::_accept, we will also
> send TASK_ERROR.
>
>
> Diffs
> -----
>
> src/master/master.cpp d94a8510c4cee9c010706f79caf27ef4a10b41a8
>
> Diff: https://reviews.apache.org/r/51317/diff/
>
>
> Testing
> -------
>
> N/A
>
>
> Thanks,
>
> Benjamin Mahler
>
>
|