Github user acvogel commented on the pull request:
https://github.com/apache/spark/pull/5109#issuecomment-86156199
The use case which led to this patch is for long-running jobs with few
slaves where we require a large master instance type. For instance, training
decision trees using MLLib requires frequently collecting back to master, so
for large datasets we must use a large master instance type. In this use case,
the economic cost is significantly increased by having master as an on-demand
instance, which is frequently 2-3 times as expensive as a spot instance.
I understand the reasonable complexity concern that @nchammas has, and
certainly defer to the maintainer's feelings on the feature vs complexity
tradeoff.
---
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]
