mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Niklas Nielsen" <...@qni.dk>
Subject Re: Review Request 38253: Add containerId to ResourceUsage to enable QoS controller to target a container
Date Thu, 17 Sep 2015 00:46:04 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/38253/#review99340
-----------------------------------------------------------



include/mesos/mesos.proto (line 820)
<https://reviews.apache.org/r/38253/#comment156232>

    The comment doesn't add a ton of context; let's kill this one or specify that it is the
container id for the executor specified in the executor_info field.



include/mesos/slave/oversubscription.proto (lines 41 - 42)
<https://reviews.apache.org/r/38253/#comment156233>

    How about something like:
    
    NOTE: Framework ID and either executor id or executor id **and** container_id must be
set to kill a running container. In the latter case, the caller can be assured that the kill
request doesn't target a new executor reusing an old executor id.



src/slave/slave.cpp (line 4380)
<https://reviews.apache.org/r/38253/#comment156234>

    Great! Let's include the container id's as well for debugging :)


- Niklas Nielsen


On Sept. 15, 2015, 6:08 a.m., Klaus Ma wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/38253/
> -----------------------------------------------------------
> 
> (Updated Sept. 15, 2015, 6:08 a.m.)
> 
> 
> Review request for mesos and Niklas Nielsen.
> 
> 
> Bugs: MESOS-2875
>     https://issues.apache.org/jira/browse/MESOS-2875
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> We should ensure that we are addressing the container which the QoS controller intended
to kill. Without this check, we may run into a scenario where the executor has terminated
and one with the same id has started in the interim i.e. running in a different container
than the one the QoS controller targeted.
> 
> This most likely requires us to add containerId to the ResourceUsage message and encode
the containerID in the QoS Correction message.
> 
> 
> Diffs
> -----
> 
>   include/mesos/mesos.proto b1deed4 
>   include/mesos/slave/oversubscription.proto fa69a95 
>   src/slave/slave.cpp 44865bd 
>   src/tests/oversubscription_tests.cpp 0c5edaf 
> 
> Diff: https://reviews.apache.org/r/38253/diff/
> 
> 
> Testing
> -------
> 
> make
> make check
> 
> 
> Thanks,
> 
> Klaus Ma
> 
>


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