> On Sept. 14, 2015, 6:20 p.m., Bill Farner wrote: > > docs/configuration-reference.md, lines 344-355 > > <https://reviews.apache.org/r/38390/diff/2/?file=1073338#file1073338line344> > > > > Tiers seems like a significant enough topic to warrant its own page > > with some more context and better flow. As it stands, it's really hard to > > understand what exactly a tier is and why i want or need to define them. > > > > Also, you might consider dropping the term 'best-effort' and stick to > > 'revocable' to avoid overloading the naming. > > Maxim Khutornenko wrote: > | Tiers seems like a significant enough topic to warrant its own page > with some more context and better flow. As it stands, it's really hard to > understand what exactly a tier is and why i want or need to define them. > > Agree. I am hesitant to promote the `tier` concept though until we fully > conceptualize it in AURORA-1443. As it stands now, it only supports > `revocable` value and has zero meaning outside the revocable offer work. > > | Also, you might consider dropping the term 'best-effort' and stick to > 'revocable' to avoid overloading the naming. > > Sure, works for me.
The intention is not to promote it (disclaimers are fine), but rather to paint the picture and not leave the reader hanging. Doesn't have to be a lot of content, but some indication of how it fits in and the development status would go a long way. > On Sept. 14, 2015, 6:20 p.m., Bill Farner wrote: > > docs/deploying-aurora-scheduler.md, line 201 > > <https://reviews.apache.org/r/38390/diff/2/?file=1073339#file1073339line201> > > > > Would it make sense to remove this flag, and instead enable revocable > > resources when there's at least one configured tier that uses them? > > Maxim Khutornenko wrote: > I don't think we want to tie these together. We should be able to > start/stop receiving revocable resources independent of the revocable tier > presence. Modifying a tier config file to stop receiving mesos revocable > offers is very confusing and error prone. I buy this for the state today, but it seems like this should change if we ever make explicit tier configuration required. The current situation is error-prone since the cluster operator needs to be sure to flip two knobs to really 'enable' the feature. - Bill ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/38390/#review98977 ----------------------------------------------------------- On Sept. 15, 2015, 5:38 p.m., Maxim Khutornenko wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/38390/ > ----------------------------------------------------------- > > (Updated Sept. 15, 2015, 5:38 p.m.) > > > Review request for Aurora and Bill Farner. > > > Bugs: AURORA-1441 > https://issues.apache.org/jira/browse/AURORA-1441 > > > Repository: aurora > > > Description > ------- > > Adding oversubscription summary. > > > Diffs > ----- > > docs/configuration-reference.md ad2701cadd38bb2fdbbe2acc477038986f8ec733 > docs/deploying-aurora-scheduler.md 8db0e615b6abe6865a889dbcfb24271655caaee6 > > Diff: https://reviews.apache.org/r/38390/diff/ > > > Testing > ------- > > Private remote: > https://github.com/maxim111333/incubator-aurora/blob/oversubscription_docs/docs/ > > > Thanks, > > Maxim Khutornenko > >
