> On Dec. 1, 2017, 12:54 a.m., Michael Park wrote: > > src/master/master.cpp > > Lines 8129-8136 (patched) > > <https://reviews.apache.org/r/64194/diff/1/?file=1904323#file1904323line8129> > > > > I found this example a bit difficult to follow... > > Maybe it's enough to link the JIRA and leave a simple > > comment saying: > > > > > "pre-MULTI_ROLE" schedulers are not `AllocationInfo` > > > aware, and since they may be performing operations > > > that implicitly uses all of `Resource`'s state > > > (e.g., equality comparison), we strip > > > the `AllocationInfo` from `Resource`, as well as > > > `Offer`.
Sounds good, thanks for the suggestion! - Benjamin ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/64194/#review192417 ----------------------------------------------------------- On Nov. 30, 2017, 1:45 a.m., Benjamin Mahler wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/64194/ > ----------------------------------------------------------- > > (Updated Nov. 30, 2017, 1:45 a.m.) > > > Review request for mesos, James DeFelice and Michael Park. > > > Bugs: MESOS-8237 > https://issues.apache.org/jira/browse/MESOS-8237 > > > Repository: mesos > > > Description > ------- > > Per MESOS-8237, it is problematic to show `Resource.allocation_info` > for "old" schedulers. The example that was presented was the > "mesos-go" 3rd party library that provides resource comparison logic. > The library has both MULTI-ROLE and non-MULTI_ROLE schedulers as > clients, these clients use the comparison logic supplied by the > library to see if the offer contains the resources they need. > The library author does not want to have to do any stripping of the > allocation info to preserve the equality. > > > Diffs > ----- > > src/master/master.cpp 700e12433b0b66efc3f5dd296711c0f203a13144 > > > Diff: https://reviews.apache.org/r/64194/diff/1/ > > > Testing > ------- > > Updated the tests in the subsequent review. > > > Thanks, > > Benjamin Mahler > >
