Hide Forgot
Description of problem: after every time migrated, offset value will add about 0.1 second. Version-Release number of selected component (if applicable): host info: # uname -r 2.6.18-266.el5 # rpm -qa|grep kvm kvm-83-236.el5 guest info: How reproducible: Steps to Reproduce: 1. (1)src host: /usr/libexec/qemu-kvm -M rhel5.4.4 -rtc-td-hack -usbdevice tablet -no-hpet -drive file=/mnt/rhel5.7-32.raw,if=virtio,boot=on,werror=stop,cache=none,format=raw,media=disk -cpu qemu64 -smp 4,sockets=2,cores=1,threads=2 -m 2G -net nic,macaddr=00:12:24:57:a8:2e,model=virtio,vlan=0 -net tap,script=/etc/qemu-ifup,vlan=0 -uuid `uuidgen` -boot c -balloon none -notify all -rtc-td-hack -no-kvm-pit-reinjection -monitor stdio -vnc :1 (2)des host: /usr/libexec/qemu-kvm -M rhel5.4.4 -rtc-td-hack -usbdevice tablet -no-hpet -drive file=/mnt/rhel5.7-32.raw,if=virtio,boot=on,werror=stop,cache=none,format=raw,media=disk -cpu qemu64 -smp 4,sockets=2,cores=1,threads=2 -m 2G -net nic,macaddr=00:12:24:57:a8:2e,model=virtio,vlan=0 -net tap,script=/etc/qemu-ifup,vlan=0 -uuid `uuidgen` -boot c -balloon none -notify all -rtc-td-hack -no-kvm-pit-reinjection -monitor stdio -vnc :1 -incoming tcp:0:5555 2.add option divider=10 lpj=n clocksource=acpi_pm to Guest kernel cmdline 3. sync two host and guest time before do migrate ntpdate -q clock.redhat.com 3. do 11 times ping-pong migration Actual results: offset value exceed 1 second Expected results: offset value always less then 1 second. Additional info:
As this is not our default clock source (KVMclock), and the problem can be taken care of by NTP, this bug will not be fixed for RHEL5.