> On Jan. 17, 2017, 5:10 p.m., Jay Guo wrote:
> > Also, should we allow user to downgrade from a multi-role framework to single-role?
I feel it would be very complicated and we should explicitly disallow that...
>
> Benjamin Bannier wrote:
> I am not sure this is required. We already make sure elsewhere that (1) only one
of `role` or `roles` can be set, and (2) we also make sure that multirole frameworks cannot
change their roles. Are there any implications of a framework opting in and out of `MULTI_ROLE`?
In either case it would be in at most one role (some role or `*`). I have the feeling that
adding such a validation might make experimenting with this feature harder than needed (if
there are no technical reasons to explicitly forbid it). Any ideas @bmahler?
I was thinking that in a longer run, is it a valid use case that user wants to downgrade framework
with `roles == {role1, role2}` to `role == role1`?
- Jay
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/55271/#review161828
-----------------------------------------------------------
On Jan. 18, 2017, 6:20 a.m., Benjamin Bannier wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/55271/
> -----------------------------------------------------------
>
> (Updated Jan. 18, 2017, 6:20 a.m.)
>
>
> Review request for mesos, Benjamin Mahler, Jay Guo, and Guangya Liu.
>
>
> Bugs: MESOS-6631
> https://issues.apache.org/jira/browse/MESOS-6631
>
>
> Repository: mesos
>
>
> Description
> -------
>
> We currently do not allow `MULTI_ROLE` frameworks to change their
> roles. This restriction will be lifted later.
>
>
> Diffs
> -----
>
> src/master/master.hpp 44f4fecb1fbe8bebf830990a59a5462338e6e004
> src/tests/master_validation_tests.cpp e5d55e03648cb218d42adc594d6fa7d40ea9bcbb
>
> Diff: https://reviews.apache.org/r/55271/diff/
>
>
> Testing
> -------
>
> Tested on various Linux configurations in internal CI.
>
>
> Thanks,
>
> Benjamin Bannier
>
>
|