----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/72314/#review220217 -----------------------------------------------------------
Ship it! Thanks for jumping on this with a fix! src/master/allocator/mesos/hierarchical.cpp Lines 2463-2464 (patched) <https://reviews.apache.org/r/72314/#comment308535> This is technically a race given the non-atomicity of a group of `addSlave + addFramework + addFramework + ...` calls, right? Perhaps note that here and maybe even a TODO to suggest making that atomic or something? src/master/allocator/mesos/hierarchical.cpp Line 2463 (original), 2469 (patched) <https://reviews.apache.org/r/72314/#comment308534> newline? - Benjamin Mahler On April 3, 2020, 12:53 p.m., Andrei Sekretenko wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/72314/ > ----------------------------------------------------------- > > (Updated April 3, 2020, 12:53 p.m.) > > > Review request for mesos, Andrei Budnik, Benjamin Mahler, and Greg Mann. > > > Bugs: MESOS-10109 > https://issues.apache.org/jira/browse/MESOS-10109 > > > Repository: mesos > > > Description > ------- > > This patch fixes MESOS-10109 by making hierarchical allocator, when > it generates inverse offers, skip IDs of frameworks returned by > `Slave::getOfferedOrAllocated()` that have not yet been added > to the allocator via `addFramework()`. > > > Diffs > ----- > > src/master/allocator/mesos/hierarchical.cpp > 0df319b6b1475cc3bb46f88e3788846ecf353468 > > > Diff: https://reviews.apache.org/r/72314/diff/1/ > > > Testing > ------- > > `make check` (+ the test from the next patch) > > > Thanks, > > Andrei Sekretenko > >
