You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@mesos.apache.org by Jeff Schroeder <je...@computer.org> on 2016/05/02 23:31:14 UTC

Status of MESOS-2533?

Some frameworks like Aurora use custom executors to distribute the
healthchecks with the tasks. This allows the task to survive a network
partition without the scheduler setting it to TASK_LOST.

Marathon uses mesos-health-check for command based health checks, but does
TCP and HTTP healthchecks from the elected scheduler (marathon issue
#3728). On a partition event, it sets those tasks to TASK_LOST causing the
master to kill them on partition heal. It also means the scheduler gets
bogged down when you have many tasks with many healthchecks defined.

Can this feature get a Shepard as would be useful for making mesos tasks
more resilient in general? There is an open review from Haosdent for fixing
it.

Thanks!


-- 
Text by Jeff, typos by iPhone

Re: Status of MESOS-2533?

Posted by Jeff Schroeder <je...@computer.org>.
Thanks for answering. Curl works fine, but that requires having curl in
your container :)

It just seems like it makes a ton of sense for mesos to perform this
natively, like your patch does.

On Wednesday, May 4, 2016, haosdent <ha...@gmail.com> wrote:

> Sorry for blocked by other things recently and didn't reply in time. @Jeff
> I have already contact Alex last week, he would review this shortly when he
> available. However, I think you could use curl to check the http status
> instead so far? Do you encounter any problems when using curl in the
> command health check?
>
> On Thu, May 5, 2016 at 1:16 AM, Benjamin Mahler <bmahler@apache.org
> <javascript:_e(%7B%7D,'cvml','bmahler@apache.org');>> wrote:
>
>> +AlexR
>>
>> On Mon, May 2, 2016 at 2:31 PM, Jeff Schroeder <
>> jeffschroeder@computer.org
>> <javascript:_e(%7B%7D,'cvml','jeffschroeder@computer.org');>> wrote:
>>
>>> Some frameworks like Aurora use custom executors to distribute the
>>> healthchecks with the tasks. This allows the task to survive a network
>>> partition without the scheduler setting it to TASK_LOST.
>>>
>>> Marathon uses mesos-health-check for command based health checks, but
>>> does TCP and HTTP healthchecks from the elected scheduler (marathon issue
>>> #3728). On a partition event, it sets those tasks to TASK_LOST causing the
>>> master to kill them on partition heal. It also means the scheduler gets
>>> bogged down when you have many tasks with many healthchecks defined.
>>>
>>> Can this feature get a Shepard as would be useful for making mesos tasks
>>> more resilient in general? There is an open review from Haosdent for fixing
>>> it.
>>>
>>> Thanks!
>>>
>>>
>>> --
>>> Text by Jeff, typos by iPhone
>>>
>>
>>
>
>
> --
> Best Regards,
> Haosdent Huang
>


-- 
Text by Jeff, typos by iPhone

Re: Status of MESOS-2533?

Posted by haosdent <ha...@gmail.com>.
Sorry for blocked by other things recently and didn't reply in time. @Jeff
I have already contact Alex last week, he would review this shortly when he
available. However, I think you could use curl to check the http status
instead so far? Do you encounter any problems when using curl in the
command health check?

On Thu, May 5, 2016 at 1:16 AM, Benjamin Mahler <bm...@apache.org> wrote:

> +AlexR
>
> On Mon, May 2, 2016 at 2:31 PM, Jeff Schroeder <jeffschroeder@computer.org
> > wrote:
>
>> Some frameworks like Aurora use custom executors to distribute the
>> healthchecks with the tasks. This allows the task to survive a network
>> partition without the scheduler setting it to TASK_LOST.
>>
>> Marathon uses mesos-health-check for command based health checks, but
>> does TCP and HTTP healthchecks from the elected scheduler (marathon issue
>> #3728). On a partition event, it sets those tasks to TASK_LOST causing the
>> master to kill them on partition heal. It also means the scheduler gets
>> bogged down when you have many tasks with many healthchecks defined.
>>
>> Can this feature get a Shepard as would be useful for making mesos tasks
>> more resilient in general? There is an open review from Haosdent for fixing
>> it.
>>
>> Thanks!
>>
>>
>> --
>> Text by Jeff, typos by iPhone
>>
>
>


-- 
Best Regards,
Haosdent Huang

Re: Status of MESOS-2533?

Posted by Benjamin Mahler <bm...@apache.org>.
+AlexR

On Mon, May 2, 2016 at 2:31 PM, Jeff Schroeder <je...@computer.org>
wrote:

> Some frameworks like Aurora use custom executors to distribute the
> healthchecks with the tasks. This allows the task to survive a network
> partition without the scheduler setting it to TASK_LOST.
>
> Marathon uses mesos-health-check for command based health checks, but does
> TCP and HTTP healthchecks from the elected scheduler (marathon issue
> #3728). On a partition event, it sets those tasks to TASK_LOST causing the
> master to kill them on partition heal. It also means the scheduler gets
> bogged down when you have many tasks with many healthchecks defined.
>
> Can this feature get a Shepard as would be useful for making mesos tasks
> more resilient in general? There is an open review from Haosdent for fixing
> it.
>
> Thanks!
>
>
> --
> Text by Jeff, typos by iPhone
>