----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/60525/#review179938 -----------------------------------------------------------
As we discussed, let's check with benM, alexR about whether or not this new behavior is appropriate. Looks good to me, but I'd like to confirm with them. src/master/allocator/mesos/hierarchical.cpp Line 1044 (original), 1044 (patched) <https://reviews.apache.org/r/60525/#comment254876> ;; src/master/allocator/mesos/hierarchical.cpp Lines 1046-1052 (patched) <https://reviews.apache.org/r/60525/#comment254880> Consider nesting this within the `seconds.isError` check, here and below. - Greg Mann On June 28, 2017, 11:25 p.m., Gastón Kleiman wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/60525/ > ----------------------------------------------------------- > > (Updated June 28, 2017, 11:25 p.m.) > > > Review request for mesos, Alexander Rukletsov, Benjamin Mahler, Greg Mann, > and Vinod Kone. > > > Bugs: MESOS-7660 > https://issues.apache.org/jira/browse/MESOS-7660 > > > Repository: mesos > > > Description > ------- > > If a framework accepts/refuses an offer using a very long filter, the > `HierarchicalAllocator` will use the default filter instead. Meaning > that it will filter the resources for only 5 seconds. This can happen > when a framework sets `Filter::refuse_seconds` to a number of seconds > larger than what fits in `Duration`. > > This patch makes the allocator use the largest possible filter duration > in this case. > > > Diffs > ----- > > src/master/allocator/mesos/hierarchical.cpp > eb01d8e6b1108866ebc049f9f4a46157823a3541 > > > Diff: https://reviews.apache.org/r/60525/diff/1/ > > > Testing > ------- > > `make check` > > > Thanks, > > Gastón Kleiman > >