----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/34970/#review86352 -----------------------------------------------------------
src/examples/oversubscribing_framework.cpp <https://reviews.apache.org/r/34970/#comment138310> Also, note that with my changes in fight, a task cannot use revocable resources if it's executor does not. This currently means a command task cannot be launched. The right fix is for the master to ignore this check for command executors and update the Slave::getExecutorInfo() set executor's resources as revocable if the task uses revocable resources. I will work on it next. - Vinod Kone On June 3, 2015, 1:21 a.m., Ben Mahler wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/34970/ > ----------------------------------------------------------- > > (Updated June 3, 2015, 1:21 a.m.) > > > Review request for mesos, Jie Yu and Vinod Kone. > > > Bugs: MESOS-2655 > https://issues.apache.org/jira/browse/MESOS-2655 > > > Repository: mesos > > > Description > ------- > > Based on a flag, this framework oversubscribes a particular resource. > > Currently, only `--oversubscribed_resource=cpus` is supported, which launches > tasks that consume revocable cpus using a tight while loop in the shell. > > > Diffs > ----- > > src/Makefile.am a5a7306b1ef65ca2b643653779ab76c26dbb5c90 > src/examples/oversubscribing_framework.cpp PRE-CREATION > > Diff: https://reviews.apache.org/r/34970/diff/ > > > Testing > ------- > > Since we do not have an non-zero estimator yet, I modified the slave to send > revocable resources in order to manually test this. > > > Thanks, > > Ben Mahler > >
