Description
It is possible to have spark apps that work best with either more memory or more CPU.
In a multi-tenant environment (such as Mesos) it can be very beneficial to be able to limit the Coarse scheduler to guarantee an executor doesn't subscribe to too many cpus or too much memory.
This ask is to add functionality to the Coarse Mesos Scheduler to have basic limits to the ratio of memory to cpu, which default to the current behavior of soaking up whatever resources it can.
Attachments
Issue Links
- relates to
-
SPARK-5095 Support launching multiple mesos executors in coarse grained mesos mode
- Resolved
- links to