> On April 19, 2018, 1:56 p.m., Qian Zhang wrote:
> > src/slave/paths.cpp
> > Line 761 (original), 763 (patched)
> > <https://reviews.apache.org/r/66705/diff/1/?file=2005759#file2005759line766>
> >
> >     Do you want to kill this `CHECK_SOME` too?

We could return the error for consistency, but we really don't expect this to 
fail. We know the target file exists, so the `rm` failing would be a 
consistency error.


> On April 19, 2018, 1:56 p.m., Qian Zhang wrote:
> > src/slave/slave.cpp
> > Lines 8891-8906 (original), 8892-8907 (patched)
> > <https://reviews.apache.org/r/66705/diff/1/?file=2005760#file2005760line8892>
> >
> >     I see you have the code below in the patch 
> > https://reviews.apache.org/r/66706 , but I think it should be part of this 
> > patch because in any cases we should not use the data of a `Try` object 
> > before making sure it has no error. Or can we just merge this patch and 
> > r66706 into one patch since they are closely related?
> >     ```
> >       if (directory.isError()) {
> >         return Error(directory.error());
> >       }
> >     ```

I want to keep these two patches separate because it makes it a lot easier to 
reason about the changes in behavior. I've added a `CHECK_SOME` to this patch 
to preserve the original behavior (until the next patch).


> On April 19, 2018, 1:56 p.m., Qian Zhang wrote:
> > src/slave/slave.cpp
> > Lines 9614 (patched)
> > <https://reviews.apache.org/r/66705/diff/1/?file=2005760#file2005760line9614>
> >
> >     Is it possible for agent to crash here?

Not in any new ways. Previously an I/O error creating the directory would cause 
the crash there. I didn't want to silently drop that error, so we would now 
crash here instead. There's sometehing fatally wrong if we can't write the 
checkpoint successfully.


- James


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/66705/#review201511
-----------------------------------------------------------


On April 18, 2018, 10:12 p.m., James Peach wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/66705/
> -----------------------------------------------------------
> 
> (Updated April 18, 2018, 10:12 p.m.)
> 
> 
> Review request for mesos, Gilbert Song, Greg Mann, and Jie Yu.
> 
> 
> Bugs: MESOS-8585
>     https://issues.apache.org/jira/browse/MESOS-8585
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Rather than crashing if the agent fails to create the executor
> directory, propagate the error to the caller so that it can
> handle it appropriately.
> 
> 
> Diffs
> -----
> 
>   src/slave/paths.hpp fe5ab9e7f96d69069406e2714ab676a5bb070534 
>   src/slave/paths.cpp 690bfe3587e6d728ab6eb712a913de23c4abe353 
>   src/slave/slave.cpp e5d6c3fac5054a6b0a0b5b77abd850a35be6ccc5 
>   src/tests/paths_tests.cpp dc765ed9db7a8ac7ca0bcb4af5cf353547ba881f 
> 
> 
> Diff: https://reviews.apache.org/r/66705/diff/1/
> 
> 
> Testing
> -------
> 
> make check (Fedora 27)
> 
> 
> Thanks,
> 
> James Peach
> 
>

Reply via email to