> On Feb. 4, 2015, 11:47 a.m., Alexander Rukletsov wrote:
> > Ben, I like the refactor you've done. A special thanks goes for updating 
> > comments for clarity and brevity. Below are my 2ยข on naming.
> > 
> > In general, the naming of executors and executor drivers in our codebase is 
> > a bit ambiguous. We have basic executor that lives in `exec.cpp` which is 
> > usually called the "executor driver", but we also have the executor driver, 
> > say `MesosExecutorDriver`, which is not relevant here. Therefore I can't 
> > decide what is the best name for the timeout that lives in `exec.cpp`, I 
> > opted for `ExecGracePeriod`, you for `ExecutorDriverGracePeriod`. Maybe we 
> > can come up with a better, less ambigouos name?
> > 
> > Again, `ExecutorGracePeriod` IMO doesn't reflect that this timeout lives in 
> > containerizer. The innermost timeout, which you renamed to 
> > `CommandExecutorGracePeriod`, can be in theory used by other "real" 
> > executors, that's why I gave it a more general name. But most probably 
> > custom executors will use the parameter from `ExecutorInfo`, so let's leave 
> > `CommandExecutorGracePeriod`.

Thanks for the feedback!

Two important notes:

> We have basic executor that lives in exec.cpp which is usually called the 
> "executor driver", but we also have the executor driver, say 
> MesosExecutorDriver, which is not relevant here.

(1) exec.cpp is not a basic executor, it's the executor driver. The executor 
driver is the library that executors leverage, think of the `ExecutorProcess` 
as being an internal part of the driver, not an executor in its own right. This 
is why I call it the `executorDriverGracePeriod`, make sense?

> Again, ExecutorGracePeriod IMO doesn't reflect that this timeout lives in 
> containerizer.

(2) It doesn't live in the containerizer, it lives in the slave :) This is why 
I call it slave::getExecutorGracePeriod.

> The innermost timeout, which you renamed to CommandExecutorGracePeriod, can 
> be in theory used by other "real" executors, that's why I gave it a more 
> general name.

(3) Are you expecting executors to have access to graceful_shutdown.hpp, or am 
I missing something?


> On Feb. 4, 2015, 11:47 a.m., Alexander Rukletsov wrote:
> > src/slave/graceful_shutdown.hpp, lines 22-23
> > <https://reviews.apache.org/r/30580/diff/1/?file=846975#file846975line22>
> >
> >     This is not necessary here: we take `Duration` as a function parameter, 
> > which means the caller already has access to `Duration`; and we do not 
> > manipulate `Duration` instances. An include in `.cpp` suffices.

We haven't been relying on transitive includes in general, that gets difficult 
to manage unless you have #include tooling.


> On Feb. 4, 2015, 11:47 a.m., Alexander Rukletsov wrote:
> > src/slave/graceful_shutdown.cpp, line 67
> > <https://reviews.apache.org/r/30580/diff/1/?file=846976#file846976line67>
> >
> >     I think a `CHECK` is too strong. It would be nice to have validation in 
> > flags loading functionality, though.

I've added a TODO to have a EXECUTOR_SHUTDOWN_GRACE_PERIOD_MAXIMUM as well, in: 
https://reviews.apache.org/r/30601/


- Ben


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/30580/#review70949
-----------------------------------------------------------


On Feb. 4, 2015, 2:14 a.m., Ben Mahler wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/30580/
> -----------------------------------------------------------
> 
> (Updated Feb. 4, 2015, 2:14 a.m.)
> 
> 
> Review request for mesos, Alexander Rukletsov, Niklas Nielsen, and Vinod Kone.
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> This is a pure cleanup to clarify the documentation and naming around 
> graceful shutdown.
> 
> Note that I also namespaced the methods differently.
> 
> 
> Diffs
> -----
> 
>   src/exec/exec.cpp aada24664dba9060a92230e25689c89852585443 
>   src/launcher/executor.cpp 1cf28f168cac6e8c7e98686a35509c2b4e052504 
>   src/slave/constants.hpp 761cfafb3b1b342af4d1dbdb2dec39a45dd62794 
>   src/slave/flags.hpp 0f6cc41d60a2e3bc2121cc438351135541ef99ba 
>   src/slave/graceful_shutdown.hpp 59f5cfba032a81b5f69c2dd1bc1d96527686127c 
>   src/slave/graceful_shutdown.cpp 04d8f091dfc5a007ac46ccc91c7ff7bfe620b524 
>   src/slave/slave.cpp a8b262174ab5c9a524db8318d3d1438cd75a702b 
>   src/tests/gc_tests.cpp 454f0974833ad5db8b504a36b010cc72c3a19751 
>   src/tests/slave_tests.cpp e7e2af63da785644f3f7e6e23607c02be962a2c6 
> 
> Diff: https://reviews.apache.org/r/30580/diff/
> 
> 
> Testing
> -------
> 
> make check
> 
> 
> Thanks,
> 
> Ben Mahler
> 
>

Reply via email to