> On 一月 25, 2017, 9:55 a.m., Guangya Liu wrote: > > @Yan, I posted some comments at https://reviews.apache.org/r/51027/ for > > this issue with some comments as: > > > > ``` > > Jacob, regaring the test failure of > > OversubscriptionTest.RescindRevocableOfferWithIncreasedRevocable, I think > > that you can disable it as > > DISABLED_RescindRevocableOfferWithIncreasedRevocable in your patch, and I > > will fix this in /r/51621 > > > > I found that with current logic, it is difficult to handle this case, as we > > cannot make sure the complete order for allocator->updateSlave and > > allocator->recoverResources, if allocator->recoverResources finished first, > > then the offer will have 3 REV resources; if allocator->updateSlave > > finished first, then the offer will have 2 REV resources. > > > > With /r/51621 , after we enable recoverResources allocate resources, we can > > always make sure we get 3 REV resources in the offer, so I propose that you > > disable the test case first and I will fix this after we enable allocate > > resources when recoverResoures. > > ``` > > > > As I will have a patch for `allocate resources when recoverResoures`, so > > how about `disable` this test first and then I can fix this when my patch > > ready in https://reviews.apache.org/r/51621/ ? > > Jiang Yan Xu wrote: > Hey Guangya I took a look at your fix in /r/51621/ but had a question, > when `recoverResources()` also triggers allocations. it's true that we don't > need `Clock::advance` to trigger a periodic allocation anymore but the race > still exists, i.e., you don't know how many `Offers` events the scheduler is > going to get, right? With my fix you won't need to care about the number of > events since you really only care about the number/amount of offers (you > eventually get). So I think this fix will work with your patch as well?
@Yan, you are right. Your fix can also help `allocate when recoverResources`, my fix in /r/51621/ is still flaky for this. Thanks. - Guangya ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/55893/#review162929 ----------------------------------------------------------- On 一月 24, 2017, 9:52 p.m., Jiang Yan Xu wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/55893/ > ----------------------------------------------------------- > > (Updated 一月 24, 2017, 9:52 p.m.) > > > Review request for mesos, Benjamin Mahler, Guangya Liu, and Jacob Janco. > > > Bugs: MESOS-6904 > https://issues.apache.org/jira/browse/MESOS-6904 > > > Repository: mesos > > > Description > ------- > > This is another test fix in the chain. > > > Diffs > ----- > > src/tests/oversubscription_tests.cpp > 22ae069ab71c82c6a6e2f5783b13ebd74a9ccf25 > > Diff: https://reviews.apache.org/r/55893/diff/ > > > Testing > ------- > > make check. > > > Thanks, > > Jiang Yan Xu > >