You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@mesos.apache.org by Connor Doyle <co...@mesosphere.io> on 2014/07/11 20:49:46 UTC
Re: Review Request 22579: Add task health check and integrate with command
executor
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/22579/#review47668
-----------------------------------------------------------
src/launcher/executor.cpp
<https://reviews.apache.org/r/22579/#comment83794>
Here, when Mesos kills an unhealthy task on the framework's behalf, it emits a StatusUpdate with a task state of TASK_KILLED. Is the `healthy` field also set to `false` in this case? It would be great to allow schedulers to know why the task was killed.
- Connor Doyle
On June 20, 2014, 10:58 p.m., Timothy Chen wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/22579/
> -----------------------------------------------------------
>
> (Updated June 20, 2014, 10:58 p.m.)
>
>
> Review request for mesos, Benjamin Hindman and Niklas Nielsen.
>
>
> Repository: mesos-git
>
>
> Description
> -------
>
> Added task health check process that is launched with the command executor if health check is configured.
> It runs configured health check command and return the status to the executor to report the task health. The executor also reports the task health status back to the scheduler.
>
> The task health process keeps internal state based on health check configuration, and determine when the task it is checking for should be killed. Currently it's based on the number of consecutive failures it observed. Once the condition meets it sends a task health status update with a kill task flag turned on, and the executor will kill the task.
>
>
> Diffs
> -----
>
> include/mesos/mesos.proto 2f6be05
> src/Makefile.am b1b7d2d
> src/common/status_utils.hpp PRE-CREATION
> src/common/type_utils.hpp 27ea4d2
> src/health-check/main.cpp PRE-CREATION
> src/launcher/executor.cpp 3d55d93
> src/master/master.cpp dcf28ad
> src/messages/messages.proto 8aecc8b
> src/slave/containerizer/external_containerizer.cpp 96c434b
> src/slave/slave.cpp ed3483f
> src/tests/health_check_tests.cpp PRE-CREATION
> src/tests/script.cpp d57fc7d
>
> Diff: https://reviews.apache.org/r/22579/diff/
>
>
> Testing
> -------
>
> Added unit tests and make check.
>
>
> Thanks,
>
> Timothy Chen
>
>
Re: Review Request 22579: Add task health check and integrate with command
executor
Posted by Timothy Chen <tn...@apache.org>.
> On July 11, 2014, 6:49 p.m., Connor Doyle wrote:
> > src/launcher/executor.cpp, line 320
> > <https://reviews.apache.org/r/22579/diff/9/?file=614436#file614436line320>
> >
> > Here, when Mesos kills an unhealthy task on the framework's behalf, it emits a StatusUpdate with a task state of TASK_KILLED. Is the `healthy` field also set to `false` in this case? It would be great to allow schedulers to know why the task was killed.
Hi Connor, that makes sense, let me create another patch for that.
- Timothy
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/22579/#review47668
-----------------------------------------------------------
On June 20, 2014, 10:58 p.m., Timothy Chen wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/22579/
> -----------------------------------------------------------
>
> (Updated June 20, 2014, 10:58 p.m.)
>
>
> Review request for mesos, Benjamin Hindman and Niklas Nielsen.
>
>
> Repository: mesos-git
>
>
> Description
> -------
>
> Added task health check process that is launched with the command executor if health check is configured.
> It runs configured health check command and return the status to the executor to report the task health. The executor also reports the task health status back to the scheduler.
>
> The task health process keeps internal state based on health check configuration, and determine when the task it is checking for should be killed. Currently it's based on the number of consecutive failures it observed. Once the condition meets it sends a task health status update with a kill task flag turned on, and the executor will kill the task.
>
>
> Diffs
> -----
>
> include/mesos/mesos.proto 2f6be05
> src/Makefile.am b1b7d2d
> src/common/status_utils.hpp PRE-CREATION
> src/common/type_utils.hpp 27ea4d2
> src/health-check/main.cpp PRE-CREATION
> src/launcher/executor.cpp 3d55d93
> src/master/master.cpp dcf28ad
> src/messages/messages.proto 8aecc8b
> src/slave/containerizer/external_containerizer.cpp 96c434b
> src/slave/slave.cpp ed3483f
> src/tests/health_check_tests.cpp PRE-CREATION
> src/tests/script.cpp d57fc7d
>
> Diff: https://reviews.apache.org/r/22579/diff/
>
>
> Testing
> -------
>
> Added unit tests and make check.
>
>
> Thanks,
>
> Timothy Chen
>
>