----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/43588/#review119257 -----------------------------------------------------------
src/tests/hierarchical_allocator_tests.cpp (line 2400) <https://reviews.apache.org/r/43588/#comment180559> Move to HierarchicalAllocatorTest. src/tests/hierarchical_allocator_tests.cpp (line 2417) <https://reviews.apache.org/r/43588/#comment180557> No allocation because there is no slaves. We have trigger allocation by `Clock::advance(flags.allocation_interval);`. src/tests/hierarchical_allocator_tests.cpp (line 2423) <https://reviews.apache.org/r/43588/#comment180564> Add var on half of agents; we can used it when add 80% resources. src/tests/hierarchical_allocator_tests.cpp (line 2438) <https://reviews.apache.org/r/43588/#comment180558> delete emepty line. src/tests/hierarchical_allocator_tests.cpp (line 2442) <https://reviews.apache.org/r/43588/#comment180565> Replace `4u` to var `10 * AGENT_RECOVERY_FACTOR - half_agents_num + 1`. So when we update the value of `AGENT_RECOVERY_FACTOR` or export it as flags, we did not need to re-calculate the number. src/tests/hierarchical_allocator_tests.cpp (line 2459) <https://reviews.apache.org/r/43588/#comment180561> Add check on whether all resources are got. @Alex/@Joerg, I'd like to confirm the behaviour of following two cases: In `::addSlave`, the slaves after 80% will offer to framework firstly; the others (< 80%) will offer to framework until next allocation interval. I think we'd better to offer ALL recovered resources when 80% slaves recovered. Another case that the counter did not update in `::removeSlave`. What's our expected behaviour on the following case? * 100 slaves * 70 slaves are added ino allocator * 60 of 70 added slaves are down * another 20 slaves are added into alloator Current behaviour, allocator will continue to offer resources based on 30 slaves because "90%" are added. We'd better to wait for timeout for this case. src/tests/hierarchical_allocator_tests.cpp (line 2507) <https://reviews.apache.org/r/43588/#comment180563> ditto src/tests/hierarchical_allocator_tests.cpp (line 2513) <https://reviews.apache.org/r/43588/#comment180566> ditto - Klaus Ma On Feb. 16, 2016, 5:23 a.m., Joerg Schad wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/43588/ > ----------------------------------------------------------- > > (Updated Feb. 16, 2016, 5:23 a.m.) > > > Review request for mesos, Alexander Rukletsov and Klaus Ma. > > > Bugs: MESOS-3986 > https://issues.apache.org/jira/browse/MESOS-3986 > > > Repository: mesos > > > Description > ------- > > Added allocator recovery tests in presence of quota. > > > Diffs > ----- > > src/tests/hierarchical_allocator_tests.cpp > 0acfc098750ff8ff9505207b983a34c1ccf3ad06 > > Diff: https://reviews.apache.org/r/43588/diff/ > > > Testing > ------- > > make check > > > Thanks, > > Joerg Schad > >