> On Aug. 27, 2015, 6:04 a.m., Alexander Rukletsov wrote:
> > include/mesos/maintenance/maintenance.proto, lines 31-37
> > <https://reviews.apache.org/r/36571/diff/15/?file=1053335#file1053335line31>
> >
> >     IIUC, here you group a bunch of machines based on time intervals. Why 
> > this grouping is necessary? We can simplify the design significantly and 
> > reduce the number of protobufs if we represent schedule by a set of 
> > maintenance events, each per machine, which may occasionally overlap in 
> > time.
> >     
> >     See my comment in https://reviews.apache.org/r/37314 with a proposal 
> > for a joined "maintenance event" message.
> 
> Joseph Wu wrote:
>     We should definitely discuss this further.
>     
>     The motivation behind this "schedule" and "window" representation is that 
> operators generally think in these terms too.  Since these protos are what 
> the operators are expected to use to interact with maintenance primitives, 
> we'd like it to match their understanding.

Absolutely, we should not surprise operators. But they do not speak protobufs, 
they will be sending JSON requests, which we may convert to something 
(protobufs) that is most convenient for us in terms of storage and retrieval.


> On Aug. 27, 2015, 6:04 a.m., Alexander Rukletsov wrote:
> > include/mesos/mesos.proto, line 136
> > <https://reviews.apache.org/r/36571/diff/15/?file=1053336#file1053336line136>
> >
> >     Why do we need this message? I've checked several following patches and 
> > haven't seen any use cases. Shall I look further up review chain?
> 
> Joseph Wu wrote:
>     We use this later on when we add the `/maintenance/start` and 
> `/maintenance/stop` endpoints, which both take a list of machines.
>     However, considering this https://reviews.apache.org/r/37826/ new review, 
> it may not be necessary for much longer.

Yeah, I would say the API should not much influence storage decisions. As long 
as we do not use it for storage, let's avoid it in protobufs.


- Alexander


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/36571/#review96642
-----------------------------------------------------------


On Aug. 27, 2015, 7:33 p.m., Joseph Wu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/36571/
> -----------------------------------------------------------
> 
> (Updated Aug. 27, 2015, 7:33 p.m.)
> 
> 
> Review request for mesos, Benjamin Hindman, Ben Mahler, Artem Harutyunyan, 
> Joris Van Remoortere, and Vinod Kone.
> 
> 
> Bugs: MESOS-3066
>     https://issues.apache.org/jira/browse/MESOS-3066
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> * MachineInfo - Describes a single box that holds one or more agents.
> * MachineInfos - A list of boxes.
> * maintenance::Window - A set of machines and a planned downtime period.
> * maintenance::Schedule - A set of maintenance windows.
> * maintenance::Mode - An enum for the three states of maintenance: Normal, 
> Draining, Deactivated.
> * Registry::MaintenanceStatus - Holds the maintenance mode of a machine.
> 
> 
> Diffs
> -----
> 
>   include/mesos/maintenance/maintenance.hpp PRE-CREATION 
>   include/mesos/maintenance/maintenance.proto PRE-CREATION 
>   include/mesos/mesos.proto 715b8cf38e1e56c18a3f2ddbb82c920bd9414f05 
>   src/Makefile.am 7b620ff66856b3f0adac121b3297d55ed71a3d99 
>   src/master/registry.proto a1995e56886f5296bc71d1bdfebec0d8316396c6 
> 
> Diff: https://reviews.apache.org/r/36571/diff/
> 
> 
> Testing
> -------
> 
> `make check`
> 
> 
> Thanks,
> 
> Joseph Wu
> 
>

Reply via email to