On Mon, Nov 25, 2013 at 4:50 PM, Doron Fediuck wrote:
Hi Gianluca, KSM comes with qemu-kvm, so no point of looking for it as an rpm: # rpm -ql qemu-kvm | grep ksm /etc/ksmtuned.conf /etc/rc.d/init.d/ksm /etc/rc.d/init.d/ksmtuned /etc/sysconfig/ksm /usr/sbin/ksmtuned
As for the actual issue, which cluster policy are you using now? See the attached to illustrate what I'm referring to.
Are you talking about RHEL 6.x? Because on my Fedora 19 hosts (despite their names containing f18...) I see this: [root@f18ovn03 ~]# ls -l /etc/sysconfig/ksm ls: cannot access /etc/sysconfig/ksm: No such file or directory [root@f18ovn03 ~]# rpm -ql qemu-kvm (contains no files) [root@f18ovn03 ~]# yumdownloader ksm [root@f18ovn03 ~]# rpm -qpl ksm-1.4.2-13.fc19.x86_64.rpm /etc/ksmtuned.conf /etc/sysconfig/ksm /lib/systemd/ksmctl /lib/systemd/system/ksm.service /lib/systemd/system/ksmtuned.service /usr/sbin/ksmtuned [root@f18ovn03 ~]# rpm -qpi ksm-1.4.2-13.fc19.x86_64.rpm Name : ksm Epoch : 2 Version : 1.4.2 Release : 13.fc19 Architecture: x86_64 Install Date: (not installed) Group : Development/Tools Size : 12035 License : GPLv2+ and LGPLv2+ and BSD Signature : RSA/SHA256, Wed 06 Nov 2013 08:02:15 PM CET, Key ID 07477e65fb4b18e6 Source RPM : qemu-1.4.2-13.fc19.src.rpm Build Date : Wed 06 Nov 2013 02:11:41 AM CET Build Host : buildvm-07.phx2.fedoraproject.org Relocations : (not relocatable) Packager : Fedora Project Vendor : Fedora Project URL : http://www.qemu.org/ Summary : Kernel Samepage Merging services Description : Kernel Samepage Merging (KSM) is a memory-saving de-duplication feature, that merges anonymous (private) pages (not pagecache ones). This package provides service files for disabling and tuning KSM. As for the current cluster config: - optimization : none (disable memory page sharing) memory balloon optimization is not checked - resilience policy : migrate virtual machines - cluster policy : none It seems to me that when I was in 3.2 there was some sort of load balancing in the sense that if I powered on in sequence 3 VM they started one on a host and one on the other. So that now I noticed sort of difference. I didn't change anything in cluster policy after upgrade to 3.3.1 Gianluca