You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@mesos.apache.org by Jiang Yan Xu <ya...@jxu.me> on 2014/01/22 01:47:50 UTC

Review Request 17156: Fixed MESOS-935: Group should tell MasterDetector "no memberships detected" when it locally times out.

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/17156/
-----------------------------------------------------------

Review request for mesos and Vinod Kone.


Bugs: MESOS-935
    https://issues.apache.org/jira/browse/MESOS-935


Repository: mesos-git


Description
-------

See summary.


Diffs
-----

  src/master/detector.hpp 8df2d165751cec7918f45ced5956ca4f79093ddc 
  src/master/master.hpp 18a6cc4330d457f55f1e7eeb952d725af95ab02a 
  src/tests/master_contender_detector_tests.cpp 9cd576f21561a86588ee0ea8a8458a9371284d47 
  src/tests/master_tests.cpp d34450bf84704b224f4e2dbc61ce100b33d14027 
  src/tests/zookeeper_tests.cpp 656afdeb763f38db329aaacee7eda7ff9966573b 
  src/zookeeper/detector.hpp 00fcf5eebfa7aaeb78b4e15be9abe3a432b021c1 
  src/zookeeper/group.cpp e21dc6f76f71bfe70bd83fa14d760f28c2cfa3d9 

Diff: https://reviews.apache.org/r/17156/diff/


Testing
-------

Added MasterZooKeeperTest.LostZooKeeperCluster
make check locally and on jenkins.


Thanks,

Jiang Yan Xu


Re: Review Request 17156: Fixed MESOS-935: Group should tell MasterDetector "no memberships detected" when it locally times out.

Posted by Jiang Yan Xu <ya...@jxu.me>.

> On Jan. 21, 2014, 9:17 p.m., Vinod Kone wrote:
> > src/zookeeper/group.cpp, lines 442-443
> > <https://reviews.apache.org/r/17156/diff/1/?file=433891#file433891line442>
> >
> >     Why do this here instead of inside expired()? Don't we want the same semantics when zookeeper session times out?
> 
> Jiang Yan Xu wrote:
>     If the expiration comes from the server, we don't need to be concerned about the group not being able to receive updates about membership changes.
>     It is when we locally time out the session that we need to worry about this group not receiving updates when leadership changes while it is partitioned from ZK or ZK goes down entirely.
>     Sound alright?

After discussing with Vinod this is now put in expired() and we locally notify watcher subscribers "empty group with no memberships" when group's session expires.


- Jiang Yan


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/17156/#review32451
-----------------------------------------------------------


On Jan. 22, 2014, 2:38 p.m., Jiang Yan Xu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/17156/
> -----------------------------------------------------------
> 
> (Updated Jan. 22, 2014, 2:38 p.m.)
> 
> 
> Review request for mesos and Vinod Kone.
> 
> 
> Bugs: MESOS-935
>     https://issues.apache.org/jira/browse/MESOS-935
> 
> 
> Repository: mesos-git
> 
> 
> Description
> -------
> 
> See summary.
> 
> 
> Diffs
> -----
> 
>   src/master/detector.hpp 8df2d165751cec7918f45ced5956ca4f79093ddc 
>   src/master/master.hpp 18a6cc4330d457f55f1e7eeb952d725af95ab02a 
>   src/tests/master_contender_detector_tests.cpp 9cd576f21561a86588ee0ea8a8458a9371284d47 
>   src/tests/master_tests.cpp d34450bf84704b224f4e2dbc61ce100b33d14027 
>   src/tests/zookeeper_tests.cpp 656afdeb763f38db329aaacee7eda7ff9966573b 
>   src/zookeeper/detector.hpp 00fcf5eebfa7aaeb78b4e15be9abe3a432b021c1 
>   src/zookeeper/group.cpp e21dc6f76f71bfe70bd83fa14d760f28c2cfa3d9 
> 
> Diff: https://reviews.apache.org/r/17156/diff/
> 
> 
> Testing
> -------
> 
> Added MasterZooKeeperTest.LostZooKeeperCluster
> make check locally and on jenkins.
> 
> 
> Thanks,
> 
> Jiang Yan Xu
> 
>


Re: Review Request 17156: Fixed MESOS-935: Group should tell MasterDetector "no memberships detected" when it locally times out.

Posted by Jiang Yan Xu <ya...@jxu.me>.

> On Jan. 21, 2014, 9:17 p.m., Vinod Kone wrote:
> > src/zookeeper/group.cpp, lines 442-443
> > <https://reviews.apache.org/r/17156/diff/1/?file=433891#file433891line442>
> >
> >     Why do this here instead of inside expired()? Don't we want the same semantics when zookeeper session times out?

If the expiration comes from the server, we don't need to be concerned about the group not being able to receive updates about membership changes.
It is when we locally time out the session that we need to worry about this group not receiving updates when leadership changes while it is partitioned from ZK or ZK goes down entirely.
Sound alright?


- Jiang Yan


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/17156/#review32451
-----------------------------------------------------------


On Jan. 21, 2014, 4:47 p.m., Jiang Yan Xu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/17156/
> -----------------------------------------------------------
> 
> (Updated Jan. 21, 2014, 4:47 p.m.)
> 
> 
> Review request for mesos and Vinod Kone.
> 
> 
> Bugs: MESOS-935
>     https://issues.apache.org/jira/browse/MESOS-935
> 
> 
> Repository: mesos-git
> 
> 
> Description
> -------
> 
> See summary.
> 
> 
> Diffs
> -----
> 
>   src/master/detector.hpp 8df2d165751cec7918f45ced5956ca4f79093ddc 
>   src/master/master.hpp 18a6cc4330d457f55f1e7eeb952d725af95ab02a 
>   src/tests/master_contender_detector_tests.cpp 9cd576f21561a86588ee0ea8a8458a9371284d47 
>   src/tests/master_tests.cpp d34450bf84704b224f4e2dbc61ce100b33d14027 
>   src/tests/zookeeper_tests.cpp 656afdeb763f38db329aaacee7eda7ff9966573b 
>   src/zookeeper/detector.hpp 00fcf5eebfa7aaeb78b4e15be9abe3a432b021c1 
>   src/zookeeper/group.cpp e21dc6f76f71bfe70bd83fa14d760f28c2cfa3d9 
> 
> Diff: https://reviews.apache.org/r/17156/diff/
> 
> 
> Testing
> -------
> 
> Added MasterZooKeeperTest.LostZooKeeperCluster
> make check locally and on jenkins.
> 
> 
> Thanks,
> 
> Jiang Yan Xu
> 
>


Re: Review Request 17156: Fixed MESOS-935: Group should tell MasterDetector "no memberships detected" when it locally times out.

Posted by Vinod Kone <vi...@gmail.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/17156/#review32451
-----------------------------------------------------------



src/master/detector.hpp
<https://reviews.apache.org/r/17156/#comment61312>

    or expires right?
    
    how about "..until the group session expires"



src/tests/master_tests.cpp
<https://reviews.apache.org/r/17156/#comment61334>

    s/disconnected/schedulerDisconnected/ for consistency?



src/tests/zookeeper_tests.cpp
<https://reviews.apache.org/r/17156/#comment61336>

    s/redetect/re-detect/



src/zookeeper/group.cpp
<https://reviews.apache.org/r/17156/#comment61335>

    Why do this here instead of inside expired()? Don't we want the same semantics when zookeeper session times out?


- Vinod Kone


On Jan. 22, 2014, 12:47 a.m., Jiang Yan Xu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/17156/
> -----------------------------------------------------------
> 
> (Updated Jan. 22, 2014, 12:47 a.m.)
> 
> 
> Review request for mesos and Vinod Kone.
> 
> 
> Bugs: MESOS-935
>     https://issues.apache.org/jira/browse/MESOS-935
> 
> 
> Repository: mesos-git
> 
> 
> Description
> -------
> 
> See summary.
> 
> 
> Diffs
> -----
> 
>   src/master/detector.hpp 8df2d165751cec7918f45ced5956ca4f79093ddc 
>   src/master/master.hpp 18a6cc4330d457f55f1e7eeb952d725af95ab02a 
>   src/tests/master_contender_detector_tests.cpp 9cd576f21561a86588ee0ea8a8458a9371284d47 
>   src/tests/master_tests.cpp d34450bf84704b224f4e2dbc61ce100b33d14027 
>   src/tests/zookeeper_tests.cpp 656afdeb763f38db329aaacee7eda7ff9966573b 
>   src/zookeeper/detector.hpp 00fcf5eebfa7aaeb78b4e15be9abe3a432b021c1 
>   src/zookeeper/group.cpp e21dc6f76f71bfe70bd83fa14d760f28c2cfa3d9 
> 
> Diff: https://reviews.apache.org/r/17156/diff/
> 
> 
> Testing
> -------
> 
> Added MasterZooKeeperTest.LostZooKeeperCluster
> make check locally and on jenkins.
> 
> 
> Thanks,
> 
> Jiang Yan Xu
> 
>


Re: Review Request 17156: Fixed MESOS-935: Group should tell MasterDetector "no memberships detected" when it locally times out.

Posted by Vinod Kone <vi...@gmail.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/17156/#review32570
-----------------------------------------------------------

Ship it!


Ship It!

- Vinod Kone


On Jan. 22, 2014, 10:38 p.m., Jiang Yan Xu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/17156/
> -----------------------------------------------------------
> 
> (Updated Jan. 22, 2014, 10:38 p.m.)
> 
> 
> Review request for mesos and Vinod Kone.
> 
> 
> Bugs: MESOS-935
>     https://issues.apache.org/jira/browse/MESOS-935
> 
> 
> Repository: mesos-git
> 
> 
> Description
> -------
> 
> See summary.
> 
> 
> Diffs
> -----
> 
>   src/master/detector.hpp 8df2d165751cec7918f45ced5956ca4f79093ddc 
>   src/master/master.hpp 18a6cc4330d457f55f1e7eeb952d725af95ab02a 
>   src/tests/master_contender_detector_tests.cpp 9cd576f21561a86588ee0ea8a8458a9371284d47 
>   src/tests/master_tests.cpp d34450bf84704b224f4e2dbc61ce100b33d14027 
>   src/tests/zookeeper_tests.cpp 656afdeb763f38db329aaacee7eda7ff9966573b 
>   src/zookeeper/detector.hpp 00fcf5eebfa7aaeb78b4e15be9abe3a432b021c1 
>   src/zookeeper/group.cpp e21dc6f76f71bfe70bd83fa14d760f28c2cfa3d9 
> 
> Diff: https://reviews.apache.org/r/17156/diff/
> 
> 
> Testing
> -------
> 
> Added MasterZooKeeperTest.LostZooKeeperCluster
> make check locally and on jenkins.
> 
> 
> Thanks,
> 
> Jiang Yan Xu
> 
>


Re: Review Request 17156: Fixed MESOS-935: Group should tell MasterDetector "no memberships detected" when it locally times out.

Posted by Jiang Yan Xu <ya...@jxu.me>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/17156/
-----------------------------------------------------------

(Updated Jan. 22, 2014, 2:38 p.m.)


Review request for mesos and Vinod Kone.


Changes
-------

Addressed Vinod's comments.


Bugs: MESOS-935
    https://issues.apache.org/jira/browse/MESOS-935


Repository: mesos-git


Description
-------

See summary.


Diffs (updated)
-----

  src/master/detector.hpp 8df2d165751cec7918f45ced5956ca4f79093ddc 
  src/master/master.hpp 18a6cc4330d457f55f1e7eeb952d725af95ab02a 
  src/tests/master_contender_detector_tests.cpp 9cd576f21561a86588ee0ea8a8458a9371284d47 
  src/tests/master_tests.cpp d34450bf84704b224f4e2dbc61ce100b33d14027 
  src/tests/zookeeper_tests.cpp 656afdeb763f38db329aaacee7eda7ff9966573b 
  src/zookeeper/detector.hpp 00fcf5eebfa7aaeb78b4e15be9abe3a432b021c1 
  src/zookeeper/group.cpp e21dc6f76f71bfe70bd83fa14d760f28c2cfa3d9 

Diff: https://reviews.apache.org/r/17156/diff/


Testing
-------

Added MasterZooKeeperTest.LostZooKeeperCluster
make check locally and on jenkins.


Thanks,

Jiang Yan Xu