----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/68239/#review206907 -----------------------------------------------------------
src/tests/containerizer/cni_isolator_tests.cpp Lines 1799 (patched) <https://reviews.apache.org/r/68239/#comment290039> What's the reason of doing a pre-test cleanup first? To prevent residual rules from the previous run? src/tests/containerizer/cni_isolator_tests.cpp Lines 1929-1932 (patched) <https://reviews.apache.org/r/68239/#comment290046> Why do we need to consume a new port ID for the container port? src/tests/containerizer/cni_isolator_tests.cpp Lines 1940-1941 (patched) <https://reviews.apache.org/r/68239/#comment290042> s/`hostPort`/`hostPorts`/ s/`containerPort`/`containerPorts`/ src/tests/containerizer/cni_isolator_tests.cpp Line 1946 (original), 1975 (patched) <https://reviews.apache.org/r/68239/#comment290048> Nits: s/`statusRunning`/`statusesRunning`/ s/`statusFinished`/`statusesFinished`/ src/tests/containerizer/cni_isolator_tests.cpp Lines 1988 (patched) <https://reviews.apache.org/r/68239/#comment290050> Nit: s/`containerId`/`containerIds`/ Also, it seems we don't use `containerId` in this test once we get them at all? src/tests/containerizer/cni_isolator_tests.cpp Lines 1957-1958 (original), 1990-1996 (patched) <https://reviews.apache.org/r/68239/#comment290047> This test waits for all `TASK_RUNNING` before any `TASK_FINISHED`. What if the first `TASK_FINISHED` is delivered before the last `TASK_RUNNING`? src/tests/containerizer/cni_isolator_tests.cpp Lines 2000-2004 (patched) <https://reviews.apache.org/r/68239/#comment290049> Is it possible to just use this loop to get the container IDs and network infos? src/tests/containerizer/cni_isolator_tests.cpp Lines 2001 (patched) <https://reviews.apache.org/r/68239/#comment290054> We could use the matcher `TaskStatusStateEq(TASK_FINISHED)` to avoid the `EXPECT_EQ` below. src/tests/containerizer/cni_isolator_tests.cpp Lines 2006-2012 (patched) <https://reviews.apache.org/r/68239/#comment290052> How about setting up expectations for `lostExecutor` instead? src/tests/containerizer/cni_isolator_tests.cpp Lines 2008 (patched) <https://reviews.apache.org/r/68239/#comment290053> Nit: s/`gcSchedule`/`gcSchedules`/ src/tests/containerizer/cni_isolator_tests.cpp Lines 1984-1987 (original), 2040-2043 (patched) <https://reviews.apache.org/r/68239/#comment290055> If we use a matcher above, and use a`vector` instead of an array we can simply do `AWAIT_READY(collect(statusesFinished))` here. But I'm fine using loop here so not opening an issue. src/tests/containerizer/cni_isolator_tests.cpp Line 2000 (original), 2045-2047 (patched) <https://reviews.apache.org/r/68239/#comment290056> We could do `AWAIT_READY(collect(gcSchedule))` if `gcSchedule` is a vector. - Chun-Hung Hsiao On Aug. 6, 2018, 8:31 p.m., Jie Yu wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/68239/ > ----------------------------------------------------------- > > (Updated Aug. 6, 2018, 8:31 p.m.) > > > Review request for mesos, Chun-Hung Hsiao and Greg Mann. > > > Repository: mesos > > > Description > ------- > > This patch updated the port mapper CNI test to launch multiple > containers concurrently. This would allow us to catch the scenarios > where multiple iptables commands are executed concurrently. > > This test fails if the fix for MESOS-9125 is not included. > > > Diffs > ----- > > src/common/values.hpp 39487b955fc1a3c963f69de66ba0da869dd3ab2e > src/tests/containerizer/cni_isolator_tests.cpp > 90d2d4103c8136d2dd883318acc135f7efca80d8 > > > Diff: https://reviews.apache.org/r/68239/diff/2/ > > > Testing > ------- > > sudo make check > > > Thanks, > > Jie Yu > >