----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/45534/#review126835 -----------------------------------------------------------
Patch looks great! Reviews applied: [45529, 45533, 45534] Passed command: export OS='ubuntu:14.04' CONFIGURATION='--verbose' COMPILER='gcc' ENVIRONMENT='GLOG_v=1 MESOS_VERBOSE=1'; ./support/docker_build.sh - Mesos ReviewBot On April 4, 2016, 12:32 p.m., Benjamin Bannier wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/45534/ > ----------------------------------------------------------- > > (Updated April 4, 2016, 12:32 p.m.) > > > Review request for mesos, Alexander Rukletsov and Ben Mahler. > > > Bugs: MESOS-4760 > https://issues.apache.org/jira/browse/MESOS-4760 > > > Repository: mesos > > > Description > ------- > > Added per-role and quota share metrics to the DRFSorter. > > > Diffs > ----- > > docs/monitoring.md 76be32842e8b437fda40c8565a34ec4f8d8dfbcc > src/Makefile.am f22ae5b3bd9336a56c802e0e51d39d6cb675caf2 > src/master/allocator/mesos/hierarchical.hpp > e979fdf60da1409d1c2d08f0e9f03cef067506dd > src/master/allocator/sorter/drf/metrics.hpp PRE-CREATION > src/master/allocator/sorter/drf/metrics.cpp PRE-CREATION > src/master/allocator/sorter/drf/sorter.hpp > f316bb5b8bfe93311ecac57198392e104b234b04 > src/master/allocator/sorter/drf/sorter.cpp > c14f9a45b9e2ea8d79bd8d2f092d313afa4cbac3 > src/master/allocator/sorter/sorter.hpp > e2338d5297e11a1ca4f6e5d72a4526aa4579610c > src/tests/hierarchical_allocator_tests.cpp > 8f78a204d296f94f515f21511710a35c33e27255 > > Diff: https://reviews.apache.org/r/45534/diff/ > > > Testing > ------- > > make check (OS X, clang trunk, not optimized) > > I did also benchmark the slowdown of the allocator due to this addition with > the benchmark allocator https://reviews.apache.org/r/44853. There I saw that > for an unoptimized build this patch adds up to 70 ms to the time needed to > query the metrics endpoint (this was for the case of 5000 slaves and 1000 > frameworks), though one could expected that an optimized build might perform > better. The numbers I got where > > `#`slaves | #frameworks | old time [us] | new time [us] | slowdown > --------|-------------|---------------|---------------|--------- > 1000 | 1 | 38980 | 23847 | 0.6 > 1000 | 50 | 27834 | 42091 | 1.5 > 1000 | 100 | 40060 | 47571 | 1.2 > 1000 | 200 | 63132 | 75806 | 1.2 > 1000 | 500 | 145170 | 171929 | 1.2 > 1000 | 1000 | 427721 | 473822 | 1.1 > 5000 | 1 | 23249 | 21426 | 0.9 > 5000 | 50 | 41032 | 36318 | 0.9 > 5000 | 100 | 43636 | 45210 | 1.0 > 5000 | 200 | 60204 | 65570 | 1.1 > 5000 | 500 | 121509 | 196894 | 1.6 > 5000 | 1000 | 449476 | 496641 | 1.1 > > > Thanks, > > Benjamin Bannier > >
