Github user dragos commented on the pull request: https://github.com/apache/spark/pull/4027#issuecomment-158686949 @tnachen I think this trade-off has been discussed in [this comment](https://github.com/apache/spark/pull/4027#issuecomment-92553493) and the following three. Since there are so many comments, here's a summary: - both standalone and Yarn are using a fixed number of executor cores, so it is more user-friendly to behave in the same way - the downside is that some CPUs wouldn't be utilized this way (example: 10 free cores, `spark.executor.cores = 3`, ==> 3 executors launched, 1 core not used) - `spark.executor.cores` is optional, so when not set we can still grab all cores. Would a `max` value make sense here? I tend to agree with @pwendell and @andrewor14 but I don't want to push back if you guys discussed this previously and changed your minds (I just went through the whole thread again and I didn't find anything). Still to do: - [ ] decide on `spark.executor.cores` or having a max value instead - [ ] one [comment](https://github.com/apache/spark/pull/4027#discussion_r27091956) that wasn't addressed, related to config names. - [ ] I still need to try this on a real Mesos cluster, won't be able to do it before Monday.
--- 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 infrastruct...@apache.org or file a JIRA ticket with INFRA. --- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org