Closed oleweidner closed 9 years ago
As commented in your previous ticket on localhost core allocation (#557), this only happens when PROFILING is enabled, and is expected behavior, as it allows us to profile certain elements of the agent w/o wasting large amounts of resources. For example, I can thus run the agent scheduler over 10k cores, w/o needing to actually burning 10k cores on XSEDE or anywhere...
I added text to the documentation in that respect.
Although I dont want to spend much more time on discussing execution on laptops, this is not what the Fork LRMS currently implements.
details sigh
Hm, but I don't have profiling enabled. Unless https://github.com/radical-cybertools/radical.pilot/commit/36ec1f98d31f15d8bbf497e1c8cab84b6c77e4c0 hasn't made it into the release yet, I don't think that this is fixed.
Launching a pilot with
cores=16
on my laptop which has only 8 cores should raise an error. Instead the pilot just starts running on 8 cores.