Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

#764 Add GPU support via --enable-features #810

Open
wants to merge 3 commits into
base: master
Choose a base branch
from

Conversation

@reneploetz
Copy link

@reneploetz reneploetz commented Feb 26, 2017

This patch adds GPU scheduling support to Chronos the same way Marathon does it:

  • you need to specify "--enable-features=gpu_resources" as parameter to enable it
  • you can specify "--mesos_task_gpu=X" where X is an Integer to always allocate X GPUs for new tasks
  • the REST API was extended to allow an optional field "gpus" alongside the usual "cpus/disk/mem" arguments to allow per-task GPU settings

The argument parsing code was adapted from Marathon (Features.scala / SchedulerConfiguration.scala) to provide the same behaviour.

This should fix #764

Signed-off-by: Rene Ploetz reneploetz@gmx.de

Copy link
Member

@brndnmtthws brndnmtthws left a comment

I'd like to see a little more in the unit tests for this, but aside from that, it looks good 👍

@@ -36,7 +36,7 @@ class MesosTaskBuilderSpec extends SpecificationWithJUnit with Mockito {

val parameters = scala.collection.mutable.ListBuffer[Parameter]()

val container = Container("dockerImage", ContainerType.DOCKER, volumes, parameters, NetworkMode.HOST, None, networks, forcePullImage = true)
Copy link
Member

@brndnmtthws brndnmtthws Mar 5, 2017

Please add a new test instead of changing this one.

Copy link
Author

@reneploetz reneploetz Mar 6, 2017

I'll change the type back to DOCKER.
Do you want some tests for the MesosTaskBuilderSpec, e.g. whether a non-zero value would be applied? And do you want the commits squashed?

Copy link

@gengmao gengmao Mar 7, 2017

@reneploetz According to http://mesos.apache.org/documentation/latest/gpu-support/, seems DOCKER type is not the right one

At the time of this writing, Nvidia GPU support is only available for tasks launched through the Mesos containerizer (i.e. no support exists for launching GPU capable tasks through the Docker containerizer)

@brndnmtthws
Copy link
Member

@brndnmtthws brndnmtthws commented Mar 8, 2017

@reneploetz let me clarify: I was suggesting you add an additional (new) test with the GPUs enabled, and leave the existing test unchanged.

@mindscratch
Copy link
Contributor

@mindscratch mindscratch commented Apr 23, 2018

What's the status of this PR? Right now, mesos-agents that have GPUs won't be offered to Chronos, which is problematic when all the agents have GPUs.

@reneploetz
Copy link
Author

@reneploetz reneploetz commented Apr 26, 2018

I'm afraid I never had time to create a pull request with proper tests as per review - feel free to do so. We were using Chronos with this patch in some product, but switched to another one a while ago.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Linked issues

Successfully merging this pull request may close these issues.

4 participants