[
https://issues.apache.org/jira/browse/MESOS-4047?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15036465#comment-15036465
]
Joseph Wu commented on MESOS-4047:
----------------------------------
Note: {{MesosContainerizerSlaveRecoveryTest.ResourceStatistics}} has similar
logic for restarting the agent, re-registering an executor, and [calling
{{MesosContainerizer::usage}}|https://github.com/apache/mesos/blob/master/src/tests/slave_recovery_tests.cpp#L3267].
But this test is stable.
The flaky test waits on:
{code}
Future<Nothing> _recover = FUTURE_DISPATCH(_, &Slave::_recover);
Future<SlaveReregisteredMessage> slaveReregisteredMessage =
FUTURE_PROTOBUF(SlaveReregisteredMessage(), _, _);
{code}
Whereas the stable test waits on:
{code}
// Set up so we can wait until the new slave updates the container's
// resources (this occurs after the executor has re-registered).
Future<Nothing> update =
FUTURE_DISPATCH(_, &MesosContainerizerProcess::update);
{code}
> MemoryPressureMesosTest.CGROUPS_ROOT_SlaveRecovery is flaky
> -----------------------------------------------------------
>
> Key: MESOS-4047
> URL: https://issues.apache.org/jira/browse/MESOS-4047
> Project: Mesos
> Issue Type: Bug
> Components: test
> Affects Versions: 0.26.0
> Environment: Ubuntu 14, gcc 4.8.4
> Reporter: Joseph Wu
> Labels: flaky, flaky-test
>
> {code:title=Output from passed test}
> [----------] 1 test from MemoryPressureMesosTest
> 1+0 records in
> 1+0 records out
> 1048576 bytes (1.0 MB) copied, 0.000430889 s, 2.4 GB/s
> [ RUN ] MemoryPressureMesosTest.CGROUPS_ROOT_SlaveRecovery
> I1202 11:09:14.319327 5062 exec.cpp:134] Version: 0.27.0
> I1202 11:09:14.333317 5079 exec.cpp:208] Executor registered on slave
> bea15b35-9aa1-4b57-96fb-29b5f70638ac-S0
> Registered executor on ubuntu
> Starting task 4e62294c-cfcf-4a13-b699-c6a4b7ac5162
> sh -c 'while true; do dd count=512 bs=1M if=/dev/zero of=./temp; done'
> Forked command at 5085
> I1202 11:09:14.391739 5077 exec.cpp:254] Received reconnect request from
> slave bea15b35-9aa1-4b57-96fb-29b5f70638ac-S0
> I1202 11:09:14.398598 5082 exec.cpp:231] Executor re-registered on slave
> bea15b35-9aa1-4b57-96fb-29b5f70638ac-S0
> Re-registered executor on ubuntu
> Shutting down
> Sending SIGTERM to process tree at pid 5085
> Killing the following process trees:
> [
> -+- 5085 sh -c while true; do dd count=512 bs=1M if=/dev/zero of=./temp; done
> \--- 5086 dd count=512 bs=1M if=/dev/zero of=./temp
> ]
> [ OK ] MemoryPressureMesosTest.CGROUPS_ROOT_SlaveRecovery (1096 ms)
> {code}
> {code:title=Output from failed test}
> [----------] 1 test from MemoryPressureMesosTest
> 1+0 records in
> 1+0 records out
> 1048576 bytes (1.0 MB) copied, 0.000404489 s, 2.6 GB/s
> [ RUN ] MemoryPressureMesosTest.CGROUPS_ROOT_SlaveRecovery
> I1202 11:09:15.509950 5109 exec.cpp:134] Version: 0.27.0
> I1202 11:09:15.568183 5123 exec.cpp:208] Executor registered on slave
> 88734acc-718e-45b0-95b9-d8f07cea8a9e-S0
> Registered executor on ubuntu
> Starting task 14b6bab9-9f60-4130-bdc4-44efba262bc6
> Forked command at 5132
> sh -c 'while true; do dd count=512 bs=1M if=/dev/zero of=./temp; done'
> I1202 11:09:15.665498 5129 exec.cpp:254] Received reconnect request from
> slave 88734acc-718e-45b0-95b9-d8f07cea8a9e-S0
> I1202 11:09:15.670995 5123 exec.cpp:381] Executor asked to shutdown
> Shutting down
> Sending SIGTERM to process tree at pid 5132
> ../../src/tests/containerizer/memory_pressure_tests.cpp:283: Failure
> (usage).failure(): Unknown container: ebe90e15-72fa-4519-837b-62f43052c913
> *** Aborted at 1449083355 (unix time) try "date -d @1449083355" if you are
> using GNU date ***
> {code}
> Notice that in the failed test, the executor is asked to shutdown when it
> tries to reconnect to the agent.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
