-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/44122/
-----------------------------------------------------------

(Updated March 1, 2016, 7:39 p.m.)


Review request for mesos, Artem Harutyunyan, Jie Yu, and Timothy Chen.


Changes
-------

Looks like that `flags.work_dir` assignment has no effect.


Bugs: MESOS-4824
    https://issues.apache.org/jira/browse/MESOS-4824


Repository: mesos


Description
-------

Based on DockerContainerizerTest.ROOT_DOCKER_RecoverOrphanedPersistentVolumes.  
This tests orphaned persistent volumes and the MesosContainerizer, with and 
without rootfs.


Diffs (updated)
-----

  src/tests/containerizer/filesystem_isolator_tests.cpp 
6a60962b4593b3521c182c7320331743ccffd4ba 
  src/tests/persistent_volume_tests.cpp 
bf19c81fbcf973d1ac27fbd42eedfd7118b7ba50 

Diff: https://reviews.apache.org/r/44122/diff/


Testing
-------

As of this patch, thise tests do not pass:

1) The `filesystem/linux` isolator recovers the persistent volume as a "known" 
orphan.  Known orphans are not cleaned up immediately.
  - We only detect rootfs mounts currently.  i.e. when the root of the mount is 
inside the agent's `<work_dir>/slaves`.
  - Persistent volume mounts have roots under `<work_dir>/volumes`.
2) When the containerizer removes this orphan, the `filesystem/linux` isolator 
has no `ContainerInfo` about it.  The isolator does not look up the mount table 
again, and the isolator ignores the cleanup.
3) Test checks the mount table and sees the mount is still present -> test 
failed.


Thanks,

Joseph Wu

Reply via email to