Github user tnachen commented on the issue:
https://github.com/apache/spark/pull/14644
@srowen Mesos also supports node labels as well (which is how constraints
is implemented in Spark framework). However GPUs are implemented as a resource
(as we want to account for # of GPUs instead of just placing a task there).
As for the config name, I just picked that to begin with. I was also
thinking we should consider having a generic config name (spark.gpus?) as I
believe it could be reused. But I wasn't sure how we like to account for this
yet as GPUs are quite different from CPUs (Mesos currently just do a integer
number of GPUs, not sharing or topology information yet). You have suggestons?
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
