> On Nov. 6, 2019, 5:11 p.m., Benno Evers wrote: > > src/master/master.cpp > > Lines 3810 (patched) > > <https://reviews.apache.org/r/71729/diff/1/?file=2171587#file2171587line3810> > > > > It doesn't seem ideal to have recursively nested calls to > > `collectauthorizations()`, even if the logic is sound it seems hard to > > reason about. > > > > Would it be possible to eliminate the branching by setting `source` to > > be `resources.popReservation()` if `source` is empty?
> Would it be possible to eliminate the branching by setting source to be > resources.popReservation() if source is empty? This would only work if we know that all resources passed to `RESERVE` are indeed reserved. Unfortunately that is not the case in the current implementation (e.g., `cpus(A):1;mem:256` would reserve only `cpus`). We need to keep support for that behavior as it is part of the APII. In the patch I put up we go from the narrower extended API (e.g., all resources passed to `RESERVE` must have identical reservations) to the wider existing API so we are good. Going from wider to narrower doesn't work, though. What I could do for the sake of readibility would be to introduce a dedicated function for the legacy behavior to avoid the self-recursion. I am not sure that would help (and might it even make harder to follow the code). WDYT? > On Nov. 6, 2019, 5:11 p.m., Benno Evers wrote: > > src/master/master.cpp > > Lines 3820 (patched) > > <https://reviews.apache.org/r/71729/diff/1/?file=2171587#file2171587line3820> > > > > Shouldn't the first `Unreserve` operation contain the original `source`? Of course. > On Nov. 6, 2019, 5:11 p.m., Benno Evers wrote: > > src/master/master.cpp > > Lines 3828 (patched) > > <https://reviews.apache.org/r/71729/diff/1/?file=2171587#file2171587line3828> > > > > Is this the same as `reserve.resources().reservations()`? Good point, this is indeed the same as `reserve.resources(0).reservations`. Using that seems to be a better idea than going strictly with symmetric approaches for `targetReservations` and `ancestorReservations` as it can help avoid a number of temporaries. > On Nov. 6, 2019, 5:11 p.m., Benno Evers wrote: > > src/master/master.cpp > > Lines 3834 (patched) > > <https://reviews.apache.org/r/71729/diff/1/?file=2171587#file2171587line3834> > > > > That look more like debug-statements rather than `INFO`-level logging? Indeed, even explicitly marked up as such with my magic string, yet still missed. - Benjamin ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/71729/#review218545 ----------------------------------------------------------- On Nov. 7, 2019, noon, Benjamin Bannier wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/71729/ > ----------------------------------------------------------- > > (Updated Nov. 7, 2019, noon) > > > Review request for mesos and Benno Evers. > > > Bugs: MESOS-9991 > https://issues.apache.org/jira/browse/MESOS-9991 > > > Repository: mesos > > > Description > ------- > > This patch adds authorization handling for `RESERVE` operations > containing `source` fields. In order to stay backwards-compatible we add > a dedicated authorization branch for such operations which under the > hood translates each removed reservation to an `UNRESERVE` operation and > every added reservation as a `RESERVE` operation where we fall back to > existing authorization code for authorization. > > > Diffs > ----- > > src/master/master.cpp e7609f361b58f9b1f0d2d5eb6037f98edcb41a56 > src/tests/master_authorization_tests.cpp > 06471aa7779d399f4474ed40db3fbcc60b8298b2 > > > Diff: https://reviews.apache.org/r/71729/diff/2/ > > > Testing > ------- > > `make check` > > > Thanks, > > Benjamin Bannier > >
