mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Anand Mazumdar" <mazumdar.an...@gmail.com>
Subject Re: Review Request 41454: Added initial draft of executor HTTP API user doc.
Date Tue, 05 Jan 2016 21:28:17 GMT


> On Jan. 5, 2016, 12:20 a.m., Vinod Kone wrote:
> > docs/executor-http-api.md, lines 52-70
> > <https://reviews.apache.org/r/41454/diff/3/?file=1167954#file1167954line52>
> >
> >     looking at just the JSON, it's not very clear that `tasks` and `updates` correspond
to unacknowledged tasks and updates. we should probably rename these fields as `unacknowledged_tasks`
and `unacknowledged_updates`. what do you think?

sgtm, modified the name of the fields as per our discussion.


> On Jan. 5, 2016, 12:20 a.m., Vinod Kone wrote:
> > docs/executor-http-api.md, line 131
> > <https://reviews.apache.org/r/41454/diff/3/?file=1167954#file1167954line131>
> >
> >     Is the retry by executor necessary? The current driver doesn't do retries for
example. Are you anticipating a scenario where the executor is connected but messages are
lost?

My bad, Fixed. This got copied over verbatim from the corresponding scheduler docs.


> On Jan. 5, 2016, 12:20 a.m., Vinod Kone wrote:
> > docs/executor-http-api.md, line 343
> > <https://reviews.apache.org/r/41454/diff/3/?file=1167954#file1167954line343>
> >
> >     what's the relation between MESOS_RETRY_INTERVAL_MAX and the duration after
which the agent deems recovery process complete (EXECUTOR_REREGISTER_TIMEOUT)? AFAICT, currently
the former is set to 15 min and the latter is 2s.

As per our discussion, I modified the name of the variable to `EXECUTOR_REGISTRATION_BACKOFF_MAX`.
This variable is equal in value to `EXECUTOR_REREGISTER_TIMEOUT`. 

It essentially denotes that upon agent recovery, the agent would wait for `EXECUTOR_REREGISTER_TIMEOUT`
before killing the executor and marking the recovery process as complete. Hence, the executor
should retry atleast once within `EXECUTOR_REGISTRATION_BACKOFF_MAX` to avoid being killed.

Also, let me know if this new name looks good to you.


- Anand


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/41454/#review112636
-----------------------------------------------------------


On Jan. 5, 2016, 9:28 p.m., Anand Mazumdar wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/41454/
> -----------------------------------------------------------
> 
> (Updated Jan. 5, 2016, 9:28 p.m.)
> 
> 
> Review request for mesos and Vinod Kone.
> 
> 
> Bugs: MESOS-4177
>     https://issues.apache.org/jira/browse/MESOS-4177
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> This is an initial draft for the user doc of Executor HTTP API. Some details might change
due to the pending executor library review chain: https://reviews.apache.org/r/41275
> 
> 
> Diffs
> -----
> 
>   docs/executor-http-api.md PRE-CREATION 
>   docs/home.md d929838206817a6c49cc2343b4de82fa085da682 
> 
> Diff: https://reviews.apache.org/r/41454/diff/
> 
> 
> Testing
> -------
> 
> https://gist.github.com/hatred/d35fdaa667203bf34a78
> 
> 
> Thanks,
> 
> Anand Mazumdar
> 
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message