Join us at engage.eucalyptus.com
Hi. We have a machine with 4 CPU core and 8 GB memory (CLC, CC, NC, Walrus on the same machine). To run more instances, we are trying to play with these parameters:
# MAX_MEM=2048
MAX_MEM="16384"
MAX_CORES="8"
It works for CPU core: the cloud are able to use 8 cores for running instances (while only 4 core are physically available). But it doesnt work for memory. The system considers it can use only 7467 MB. We try to use MAX_MEM="16384" and MAX_MEM=16384 (without quotes). The same result in cc.log:
[Thu Jul 30 13:00:30 2009][005623][EUCAINFO ] node=os111yk mem=7647/7647 disk=57198/57198 cores=8/8
So our question is the following: is it possible for the cloud to use more memory than physically available (like using more CPU cores)? In other words, is it possible to share memory between running instances?
Thanks,
Yuri.
Hello,
as you noticed, no it is not possible. We don't allow the node to overcommit on the available memory at the moment.
cheers
graziano