> On Oct. 9, 2019, 1:18 p.m., Andrei Sekretenko wrote: > > src/master/quota.cpp > > Lines 235 (patched) > > <https://reviews.apache.org/r/71595/diff/1/?file=2168783#file2168783line235> > > > > Not entirely sure that we can. I've looked into the bunch of > > bugs/problems that were fixed/silenced by using the "fixed-point" > > arithmetic... > > > > Given that our "fixed point" has only 53 significant bits (as it is > > built on top of `double`), those things might start to break again at > > around 8 exabytes.
Where did you get 8 exabytes from? 9,223,372,036,854,774 megabytes is 9223 exabytes? - Benjamin ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/71595/#review218155 ----------------------------------------------------------- On Oct. 9, 2019, 4:43 a.m., Benjamin Mahler wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/71595/ > ----------------------------------------------------------- > > (Updated Oct. 9, 2019, 4:43 a.m.) > > > Review request for mesos, Andrei Sekretenko and Meng Zhu. > > > Bugs: MESOS-10008 > https://issues.apache.org/jira/browse/MESOS-10008 > > > Repository: mesos > > > Description > ------- > > Per MESOS-10008, large scalar values overflow and can crash > the master. This patch imposes maximums for guarantees and limits > well below the overflow point, to prevent this from happening. > > > Diffs > ----- > > src/master/quota.cpp 4ecd3269ac092b187082675e723cb18e40cea573 > > > Diff: https://reviews.apache.org/r/71595/diff/1/ > > > Testing > ------- > > Test added in subsequent patch. > > > Thanks, > > Benjamin Mahler > >
