You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@mesos.apache.org by Dominic Hamon <dh...@twopensource.com> on 2014/04/24 19:23:09 UTC

Review Request 20673: Moved Registrar Gauge out of internal struct.

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/20673/
-----------------------------------------------------------

Review request for mesos, Ben Mahler and Vinod Kone.


Repository: mesos-git


Description
-------

see summary.

this fixes the MasterTest.MetricsInStatsEndpoint failure. There seems to be something in the tests that causes the RegistrarProcess and internal Gauge to get out of sync. I don't have a full understanding of the issue yet, but this fixes the immediate problem.


Diffs
-----

  src/master/registrar.cpp 8e7760b2abd5f4a80a8a4b1b4b90eefc1136fe66 

Diff: https://reviews.apache.org/r/20673/diff/


Testing
-------

make check


Thanks,

Dominic Hamon


Re: Review Request 20673: Moved Registrar Gauge out of internal struct.

Posted by Dominic Hamon <dh...@twopensource.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/20673/
-----------------------------------------------------------

(Updated April 24, 2014, 10:34 a.m.)


Review request for mesos, Benjamin Hindman, Ben Mahler, and Vinod Kone.


Changes
-------

added more to description


Repository: mesos-git


Description (updated)
-------

see summary.

without this patch, running 'make check --gtest_filter="MasterTest*"' fails 100% in MasterTest.MetricsInStatsEndpoint. With this patch, this passes 100%.

Running MasterTest.MetricsInStatsEndpoint in isolation passes with or without this patch.

This leads me to conclude that there's something in the way we setup and tear down master tests that is racy. I think that we are somehow ending up with a gauge in the metrics system that is trying to call a deferred object that references a dangling registrar process reference. My theory right now is that the Gauges struct is being destroyed but the MetricsProcess removal has not run before we hit the endpoint.


Diffs
-----

  src/master/registrar.cpp 8e7760b2abd5f4a80a8a4b1b4b90eefc1136fe66 

Diff: https://reviews.apache.org/r/20673/diff/


Testing
-------

make check


Thanks,

Dominic Hamon


Re: Review Request 20673: Moved Registrar Gauge out of internal struct.

Posted by Dominic Hamon <dh...@twopensource.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/20673/
-----------------------------------------------------------

(Updated April 24, 2014, 5:29 p.m.)


Review request for mesos, Benjamin Hindman, Ben Mahler, and Vinod Kone.


Changes
-------

adding @benh to reviewers -- @vinodkone


Repository: mesos-git


Description
-------

see summary.

this fixes the MasterTest.MetricsInStatsEndpoint failure. There seems to be something in the tests that causes the RegistrarProcess and internal Gauge to get out of sync. I don't have a full understanding of the issue yet, but this fixes the immediate problem.


Diffs
-----

  src/master/registrar.cpp 8e7760b2abd5f4a80a8a4b1b4b90eefc1136fe66 

Diff: https://reviews.apache.org/r/20673/diff/


Testing
-------

make check


Thanks,

Dominic Hamon