You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@mesos.apache.org by Anand Mazumdar <an...@apache.org> on 2017/02/03 04:45:53 UTC
Review Request 56269: Introduced a `Container` struct on the default
executor.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/56269/
-----------------------------------------------------------
Review request for mesos and Vinod Kone.
Bugs: MESOS-6296
https://issues.apache.org/jira/browse/MESOS-6296
Repository: mesos
Description
-------
This helps us in consolidating various metadata associatd with
an active child container. Also, this would be used in the future
for figuring out the task group corresponding to a child container
and then killing the other child containers belonging to the task
group when one of the tasks fail; or if a task in the task group
is explicitly killed by the scheduler.
Diffs
-----
src/launcher/default_executor.cpp 97eee05cac8cb1f62d43e2aecc08a8e54e49eac3
Diff: https://reviews.apache.org/r/56269/diff/
Testing
-------
make check
Thanks,
Anand Mazumdar
Re: Review Request 56269: Introduced a `Container` struct on the
default executor.
Posted by Anand Mazumdar <an...@apache.org>.
> On Feb. 3, 2017, 11:02 p.m., Vinod Kone wrote:
> > src/launcher/default_executor.cpp, line 89
> > <https://reviews.apache.org/r/56269/diff/1/?file=1622928#file1622928line89>
> >
> > hmm. it's unfortunate that each task/container stores TaskInfos of all tasks in the task group.
> >
> > can we have a helper function instead that returns a list of container IDs that belong to the same task group as a given container?
> >
> > ```
> > hashset<ContainerID> taskGroupContainers(const ContainerID& containerId) {}
> > ```
> >
> > or store a shared pointer to TaskGroup?
We would only have a few limited active task groups at a time. So, shouldn't be that bad. This is also much easier to reason about.
Otherwise, we would need to store the taskGroups in a separate hashmap and then clean them up when all the tasks in a task group terminate among other things.
- Anand
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/56269/#review164181
-----------------------------------------------------------
On Feb. 4, 2017, 5:51 p.m., Anand Mazumdar wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/56269/
> -----------------------------------------------------------
>
> (Updated Feb. 4, 2017, 5:51 p.m.)
>
>
> Review request for mesos and Vinod Kone.
>
>
> Bugs: MESOS-6296
> https://issues.apache.org/jira/browse/MESOS-6296
>
>
> Repository: mesos
>
>
> Description
> -------
>
> This helps us in consolidating various metadata associatd with
> an active child container. Also, this would be used in the future
> for figuring out the task group corresponding to a child container
> and then killing the other child containers belonging to the task
> group when one of the tasks fail; or if a task in the task group
> is explicitly killed by the scheduler.
>
>
> Diffs
> -----
>
> src/launcher/default_executor.cpp 97eee05cac8cb1f62d43e2aecc08a8e54e49eac3
>
> Diff: https://reviews.apache.org/r/56269/diff/
>
>
> Testing
> -------
>
> make check
>
>
> Thanks,
>
> Anand Mazumdar
>
>
Re: Review Request 56269: Introduced a `Container` struct on the
default executor.
Posted by Vinod Kone <vi...@gmail.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/56269/#review164181
-----------------------------------------------------------
src/launcher/default_executor.cpp (line 89)
<https://reviews.apache.org/r/56269/#comment235804>
hmm. it's unfortunate that each task/container stores TaskInfos of all tasks in the task group.
can we have a helper function instead that returns a list of container IDs that belong to the same task group as a given container?
```
hashset<ContainerID> taskGroupContainers(const ContainerID& containerId) {}
```
or store a shared pointer to TaskGroup?
src/launcher/default_executor.cpp (line 824)
<https://reviews.apache.org/r/56269/#comment235808>
looks like we are better off keying `containers` on `TaskID` rather than `ContainerID`?
src/launcher/default_executor.cpp (lines 1033 - 1046)
<https://reviews.apache.org/r/56269/#comment235806>
Can we consolidate killed/killing into the `Container` struct as well? I think it will be easy to reason about.
- Vinod Kone
On Feb. 3, 2017, 4:45 a.m., Anand Mazumdar wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/56269/
> -----------------------------------------------------------
>
> (Updated Feb. 3, 2017, 4:45 a.m.)
>
>
> Review request for mesos and Vinod Kone.
>
>
> Bugs: MESOS-6296
> https://issues.apache.org/jira/browse/MESOS-6296
>
>
> Repository: mesos
>
>
> Description
> -------
>
> This helps us in consolidating various metadata associatd with
> an active child container. Also, this would be used in the future
> for figuring out the task group corresponding to a child container
> and then killing the other child containers belonging to the task
> group when one of the tasks fail; or if a task in the task group
> is explicitly killed by the scheduler.
>
>
> Diffs
> -----
>
> src/launcher/default_executor.cpp 97eee05cac8cb1f62d43e2aecc08a8e54e49eac3
>
> Diff: https://reviews.apache.org/r/56269/diff/
>
>
> Testing
> -------
>
> make check
>
>
> Thanks,
>
> Anand Mazumdar
>
>
Re: Review Request 56269: Introduced a `Container` struct on the
default executor.
Posted by Vinod Kone <vi...@gmail.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/56269/#review164420
-----------------------------------------------------------
Ship it!
Ship It!
- Vinod Kone
On Feb. 4, 2017, 5:51 p.m., Anand Mazumdar wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/56269/
> -----------------------------------------------------------
>
> (Updated Feb. 4, 2017, 5:51 p.m.)
>
>
> Review request for mesos and Vinod Kone.
>
>
> Bugs: MESOS-6296
> https://issues.apache.org/jira/browse/MESOS-6296
>
>
> Repository: mesos
>
>
> Description
> -------
>
> This helps us in consolidating various metadata associatd with
> an active child container. Also, this would be used in the future
> for figuring out the task group corresponding to a child container
> and then killing the other child containers belonging to the task
> group when one of the tasks fail; or if a task in the task group
> is explicitly killed by the scheduler.
>
>
> Diffs
> -----
>
> src/launcher/default_executor.cpp 97eee05cac8cb1f62d43e2aecc08a8e54e49eac3
>
> Diff: https://reviews.apache.org/r/56269/diff/
>
>
> Testing
> -------
>
> make check
>
>
> Thanks,
>
> Anand Mazumdar
>
>
Re: Review Request 56269: Introduced a `Container` struct on the
default executor.
Posted by Anand Mazumdar <an...@apache.org>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/56269/
-----------------------------------------------------------
(Updated Feb. 4, 2017, 5:51 p.m.)
Review request for mesos and Vinod Kone.
Changes
-------
Review comments.
Bugs: MESOS-6296
https://issues.apache.org/jira/browse/MESOS-6296
Repository: mesos
Description
-------
This helps us in consolidating various metadata associatd with
an active child container. Also, this would be used in the future
for figuring out the task group corresponding to a child container
and then killing the other child containers belonging to the task
group when one of the tasks fail; or if a task in the task group
is explicitly killed by the scheduler.
Diffs (updated)
-----
src/launcher/default_executor.cpp 97eee05cac8cb1f62d43e2aecc08a8e54e49eac3
Diff: https://reviews.apache.org/r/56269/diff/
Testing
-------
make check
Thanks,
Anand Mazumdar