----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/29083/#review65267 -----------------------------------------------------------
src/master/drf_sorter.hpp <https://reviews.apache.org/r/29083/#comment108345> Jie and I were chatting and realized that since the sorter aggregates resources across slaves (which breaks for non-scalar things like ranges and sets), the disk transformation would have issues with duplicated IDs here. Jie is updating the transformation to be simpler and non-idempotent which will fix the issue here (can have duplicate persistence IDs in the disk resources), but ideally we stop merging resources across slaves. At the very least a TODO is needed here! - Ben Mahler On Dec. 16, 2014, 6:28 a.m., Ben Mahler wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/29083/ > ----------------------------------------------------------- > > (Updated Dec. 16, 2014, 6:28 a.m.) > > > Review request for mesos and Jie Yu. > > > Bugs: MESOS-2099 > https://issues.apache.org/jira/browse/MESOS-2099 > > > Repository: mesos-git > > > Description > ------- > > In order for frameworks to perform resource transformations on their > allocations, the allocator must be able to update the allocation in the > sorter objects. > > Note that the sorters cannot simply take a Transformation, as done in > r/29084/. This is because sorters may not store the complete allocation for a > client (e.g. role sorter doesn't track reserved resources). > > This means that taking a Transformation is problematic for the role sorter in > the allocator. > > > Diffs > ----- > > src/master/drf_sorter.hpp c47b56afdb8f9f3ff0793f0d10a4e4c656e6a212 > src/master/drf_sorter.cpp 0ad6c52759e1ceec4b43f99d8fa1e75d7e2e1c31 > src/master/sorter.hpp 0150f90b9aab2ecb4edd07dca7f0d58997e422aa > src/tests/sorter_tests.cpp 0516ab573d9f4f2af249978e15ebf52c2afb5359 > > Diff: https://reviews.apache.org/r/29083/diff/ > > > Testing > ------- > > Added a test. > > > Thanks, > > Ben Mahler > >
