> On June 26, 2014, 1:39 a.m., Adam B wrote: > > include/mesos/mesos.proto, line 128 > > <https://reviews.apache.org/r/22123/diff/1/?file=601126#file601126line128> > > > > Please add some documentation to the FrameworkInfo comment that > > explains what a value of failover=true means and when it should be used. > > Isabel Jimenez wrote: > There is a comment about this on > https://issues.apache.org/jira/browse/MESOS-1118, do you have something in > mind ?
I'm just imagining that I'm a framework author looking at mesos.proto. I can see that it defaults to false, but I'd like to see a sentence or two in the comment above FrameworkInfo that tells me when I should set failover=true. I still don't entirely understand what problem we're trying to solve. Please correct me if I'm wrong, but from what I can read in the code, when I first register my framework, I would set failover=false and failover_timeout=X (original default) to deactivate the framework on exit/disconnect and wait X before removing it completely. I would set failover=false and failover_timeout=0 to immediately remove the framework upon exit/disconnect. I would set failover=true (timeout irrelevant) to deactivate on exit/disconnect and never removeFramework. According to MESOS-703, this timeout/flag cannot be changed during re-registrations yet. This two-flag state is really confusing. I think I prefer the suggestion that negative values (since '0' is apparently valid) imply a "failover forever" state. But you would also have to enable negative Durations https://github.com/apache/mesos/blob/master/3rdparty/libprocess/3rdparty/stout/include/stout/duration.hpp#L33 Maybe Duration::max() is enough to imply "forever"? Or if you do use two flags, I think "failover_forever=true" would read better. - Adam ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/22123/#review46725 ----------------------------------------------------------- On June 1, 2014, 9:14 p.m., Isabel Jimenez wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/22123/ > ----------------------------------------------------------- > > (Updated June 1, 2014, 9:14 p.m.) > > > Review request for mesos, Benjamin Hindman, Dominic Hamon, and Till Toenshoff. > > > Bugs: MESOS-1118 > https://issues.apache.org/jira/browse/MESOS-1118 > > > Repository: mesos-git > > > Description > ------- > > I think the name of the boolean is a bit confusing, I could change it into > 'nofailover' which I think to be clearer. > > > Diffs > ----- > > include/mesos/mesos.proto 82388e1 > src/master/master.cpp 766a0e3 > > Diff: https://reviews.apache.org/r/22123/diff/ > > > Testing > ------- > > make check > > > Thanks, > > Isabel Jimenez > >
