Re: Review Request 45386: Fix container destroy provisioning race.

[Gilbert Song]

---
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45386/
---

(Updated March 28, 2016, 10:16 a.m.)


Review request for mesos and Jie Yu.


Bugs: MESOS-5023
https://issues.apache.org/jira/browse/MESOS-5023


Repository: mesos


Description (updated)
---

Fix container destroy provisioning race.


Diffs (updated)
-

  src/slave/containerizer/mesos/containerizer.cpp 
e7f7e7fd1304e14dbfaab8b53cea16efc0417911 

Diff: https://reviews.apache.org/r/45386/diff/


Testing
---

make check

./bin/mesos-tests.sh --gtest_filter="*DestroyWhileProvisioning*" 
--gtest_repeat=-1 --gtest_break_on_failure


Thanks,

Gilbert Song

Re: Review Request 45386: Fix container destroy provisioning race.

[Jie Yu]

---
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45386/#review125664
---




src/slave/containerizer/mesos/containerizer.cpp (lines 747 - 751)


Can you adjust the comments here?


- Jie Yu


On March 28, 2016, 4:09 p.m., Gilbert Song wrote:
> 
> ---
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/45386/
> ---
> 
> (Updated March 28, 2016, 4:09 p.m.)
> 
> 
> Review request for mesos and Jie Yu.
> 
> 
> Bugs: MESOS-5023
> https://issues.apache.org/jira/browse/MESOS-5023
> 
> 
> Repository: mesos
> 
> 
> Description
> ---
> 
> Fix container destroy provisioning race.
> 
> This flaky test results from `_launch()` and `___destroy()` are waiting for 
> the same future to be ready. It is possible that `___destroy()` and its 
> dependencies already finish before `_launch()` starts. We have to check the 
> hashmap. If the `containerId` does not exist, it means destroy finishes.
> 
> 
> Diffs
> -
> 
>   src/slave/containerizer/mesos/containerizer.cpp 
> e7f7e7fd1304e14dbfaab8b53cea16efc0417911 
> 
> Diff: https://reviews.apache.org/r/45386/diff/
> 
> 
> Testing
> ---
> 
> make check
> 
> ./bin/mesos-tests.sh --gtest_filter="*DestroyWhileProvisioning*" 
> --gtest_repeat=-1 --gtest_break_on_failure
> 
> 
> Thanks,
> 
> Gilbert Song
> 
>

Re: Review Request 45386: Fix container destroy provisioning race.

[Gilbert Song]

---
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45386/
---

(Updated March 28, 2016, 9:09 a.m.)


Review request for mesos and Jie Yu.


Bugs: MESOS-5023
https://issues.apache.org/jira/browse/MESOS-5023


Repository: mesos


Description (updated)
---

Fix container destroy provisioning race.

This flaky test results from `_launch()` and `___destroy()` are waiting for the 
same future to be ready. It is possible that `___destroy()` and its 
dependencies already finish before `_launch()` starts. We have to check the 
hashmap. If the `containerId` does not exist, it means destroy finishes.


Diffs (updated)
-

  src/slave/containerizer/mesos/containerizer.cpp 
e7f7e7fd1304e14dbfaab8b53cea16efc0417911 

Diff: https://reviews.apache.org/r/45386/diff/


Testing
---

make check

./bin/mesos-tests.sh --gtest_filter="*DestroyWhileProvisioning*" 
--gtest_repeat=-1 --gtest_break_on_failure


Thanks,

Gilbert Song

Review Request 45386: Fix container destroy provisioning race.

[Gilbert Song]

---
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45386/
---

Review request for mesos and Jie Yu.


Bugs: MESOS-5023
https://issues.apache.org/jira/browse/MESOS-5023


Repository: mesos


Description
---

Fix container destroy provisioning race.

The flaky test results from `_launch()` and `___destroy()` are waiting for the 
same future to be ready. It is possible that once the future is ready and send 
callbacks to the two methods above, `___destroy()` and its dependencies finish 
before `_lauch()` starts, which erase the `containerId` in the hashmap.


Diffs
-

  src/slave/containerizer/mesos/containerizer.cpp 
e7f7e7fd1304e14dbfaab8b53cea16efc0417911 

Diff: https://reviews.apache.org/r/45386/diff/


Testing
---

make check

./bin/mesos-tests.sh --gtest_filter="*DestroyWhileProvisioning*" 
--gtest_repeat=-1 --gtest_break_on_failure


Thanks,

Gilbert Song