Package: linux-image-2.6.26-1-xen-amd64 Version: 2.6.26-13 Severity: important
When I migrate a domain from an host to another (which both run the 2.6.26-1-xen-amd64 kernel), with the packaged xen 3.2.1-2, the migrated domU have his time shifted by the timedelta of the 2 dom0 uptime. # Demo : xen-8:~# uname -a Linux xen-8 2.6.26-1-xen-amd64 #1 SMP Sat Jan 10 20:39:26 UTC 2009 x86_64 GNU/Linux xen-8:~# xm list Name ID Mem VCPUs State Time(s) Domain-0 0 500 8 r----- 2559.1 flashstream-001 86 250 2 -b---- 6.9 xen-5:~# uptime 12:36:53 up 12 days, 15 min, 1 user, load average: 0.00, 0.00, 0.00 xen-5:~# xm list Name ID Mem VCPUs State Time(s) Domain-0 0 500 8 r----- 448967.8 xen-5:~# uname -a Linux xen-5 2.6.26-1-xen-amd64 #1 SMP Sat Jan 10 20:39:26 UTC 2009 x86_64 GNU/Linux # Everyone is synced ... xen-8:~# date Tue Feb 24 12:37:17 CET 2009 xen-5:~# date Tue Feb 24 12:37:18 CET 2009 flashstream-001:~# date Tue Feb 24 12:37:19 CET 2009 # let's migrate the domU from xen-8 to xen-5 xen-8:/etc/xen# xm migrate flashstream-001 -l 10.20.0.5 On the domU : flashstream-001:~# date Tue Feb 24 12:41:35 CET 2009 # Migration is here flashstream-001:~# date Sat Mar 7 15:13:05 CET 2009 I don't encounter the bug when I use the 2.6.18-6-xen kernel from etch. I tryed to do "echo 1 > /proc/sys/xen/independent_wallclock" everywhere it's possible (in dom0 or domU), it don't change anything. Setting the clocksource to jiffies in the domU fix the problem, but it have a resolution of 4ms which sux :(. -- System Information: Debian Release: 5.0 APT prefers stable APT policy: (500, 'stable') Architecture: amd64 (x86_64) Shell: /bin/sh linked to /bin/bash Kernel: Linux 2.6.26-1-xen-amd64 Locale: lang=fr...@euro, lc_ctype=fr...@euro (charmap=ISO-8859-15) -- To UNSUBSCRIBE, email to debian-kernel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
