Description of problem: Remote migration causes a drift in guests, proportional to the difference they have in monotonic clock. This happens because msr values are not saved on migration, meaning we are just relying on bogus data for clock calculations
For reference, the Fedora 12 bug is #530389
*** Bug 523457 has been marked as a duplicate of this bug. ***
Test this in Intel host, it does not exist. kvm version: kvm-83-140.el5 hostA : Fri Jan 8 08:18:26 UTC 2010 hostB : Fri Jan 8 08:17:52 UTC 2010 1. guest: 32bitrhel5.4 kernel: 2.6.18-164.11.1.el5 cmd: /usr/libexec/qemu-kvm -rtc-td-hack -no-hpet -usbdevice tablet -no-kvm-pit-reinjection -cpu qemu64,+sse2 -drive file=RHEL-Server-5.4-32-virtio.qcow2,if=ide -smp 2 -m 2G -net nic,macaddr=20:20:20:11:12:56,model=e1000,vlan=0 -net tap,script=/etc/qemu-ifup,vlan=0 -monitor stdio -vnc :1 1.1 migrate A-->B before migration: ssh hostA date -u; date -u Fri Jan 8 08:35:49 UTC 2010 Fri Jan 8 08:35:47 UTC 2010 after migration: ssh hostA date -u; date -u Fri Jan 8 08:38:12 UTC 2010 Fri Jan 8 08:38:10 UTC 2010 1.2 migrate B-->A before migration: ssh hostB date -u; date -u Fri Jan 8 08:44:44 UTC 2010 Fri Jan 8 08:45:18 UTC 2010 after migration: ssh hostB date -u; date -u Fri Jan 8 08:47:26 UTC 2010 Fri Jan 8 08:48:00 UTC 2010 2. guest: 64bitrhel5.4 kernel: 2.6.18-164.11.1.el5 cmd: /usr/libexec/qemu-kvm -rtc-td-hack -no-hpet -usbdevice tablet -no-kvm-pit-reinjection -cpu qemu64,+sse2 -drive file=RHEL-Server-5.4-64-virtio.qcow2,if=ide -smp 2 -m 2G -net nic,macaddr=20:20:20:11:12:58,model=e1000,vlan=0 -net tap,script=/etc/qemu-ifup,vlan=0 -monitor stdio -vnc :1 2.1 Migrate A-->B before migration: ssh hostA date -u; date -u Fri Jan 8 09:29:51 UTC 2010 Fri Jan 8 09:29:50 UTC 2010 after migration: ssh hostA date -u; date -u Fri Jan 8 09:32:49 UTC 2010 Fri Jan 8 09:32:48 UTC 2010 2.2 migrate B-->A before migration: ssh hostB date -u; date -u Fri Jan 8 09:38:05 UTC 2010 Fri Jan 8 09:38:39 UTC 2010 after migration: ssh hostB date -u; date -u Fri Jan 8 09:40:15 UTC 2010 Fri Jan 8 09:40:49 UTC 2010 From the test result, no big time jump and time backward.
Test this in AMD host, it does not exist. kvm version: kvm-83-137.el5 1. guest: rhel5.4-64 guest kernel: 2.6.18-164.6.1.el5 cmd: /usr/libexec/qemu-kvm -drive file=/mnt/images/rhel5.4-64.qcow2,if=virtio,boot=on -usbdevice tablet -no-kvm-pit-reinjection -no-hpet -rtc-td-hack -smp 2 -m 2G -uuid `uuidgen` -net nic,model=virtio,macaddr=00:A6:68:BA:90:95,vlan=0 -net tap,vlan=0,script=/etc/qemu-ifup -cpu qemu64,+sse2 -boot c -name 5.4-64 -vnc :8 -monitor stdio on guest: #ssh host date -u;date -u 1.1 migrate A->B before migration: Tue Dec 8 03:31:24 UTC 2009 Tue Dec 8 03:31:26 UTC 2009 after migration: Tue Dec 8 03:35:08 UTC 2009 Tue Dec 8 03:35:10 UTC 2009 1.2 migrate B->A before migration: Thu Feb 12 23:15:34 UTC 2009 Thu Feb 12 23:15:36 UTC 2009 after migration: Thu Feb 12 23:20:20 UTC 2009 Thu Feb 12 23:20:22 UTC 2009 2. guest: rhel5.4-32 guest kernel: 2.6.18-164.6.1.el5 cmd: /usr/libexec/qemu-kvm -drive file=/mnt/images/rhel5.4-32.qcow2,if=virtio,boot=on -usbdevice tablet -no-kvm-pit-reinjection -no-hpet -rtc-td-hack -smp 2 -m 2G -uuid `uuidgen` -net nic,model=virtio,macaddr=00:A6:68:BA:90:95,vlan=0 -net tap,vlan=0,script=/etc/qemu-ifup -cpu qemu64,+sse2 -boot c -name 5.4-64 -vnc :8 -monitor stdio 2.1 migrate A->B before migration Tue Dec 8 05:58:27 UTC 2009 Tue Dec 8 06:00:05 UTC 2009 after migration: Tue Dec 8 06:01:29 UTC 2009 Tue Dec 8 06:03:06 UTC 2009 2.2 migrate B->A before migration: Tue Dec 8 05:52:08 UTC 2009 Tue Dec 8 05:51:43 UTC 2009 after migration: Tue Dec 8 05:55:24 UTC 2009 Tue Dec 8 05:54:59 UTC 2009 From the test result, no big time jump and time backward before&after migration.
*** Bug 510193 has been marked as a duplicate of this bug. ***
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on therefore solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHSA-2010-0271.html
*** Bug 578734 has been marked as a duplicate of this bug. ***