My dom0 systems all run ntpd to keep their clock synchronized, and I just
noticed my domU systems all have a little different clock, and that happens
right after boot. IE, right after xm create domU, I did a
quick''n''dirty check this way:
luca@dom0 ~ $ ssh domU "uptime; date"; date
Password:
16:11:31 up 0 min, 0 users, load average: 0.43, 0.12, 0.04
Thu Dec 8 16:11:31 CET 2011
Thu Dec 8 16:09:57 CET 2011
My systems are running Xen-4.1.1, dom0 is using Xen Linux 2.6.38
(gentoo''s xen-sources, back ported from Suse), domU is using Linux
3.0.11.
Everything is fully 64bit, in dom0 there''s xen.independent_wallclock =
0, in domU there are no *xen* sysctls.
All domains are paravirtualized, not HVM. Both dom0 and domU show
''Switching to clocksource xen'' in dmesg.
I don''t know where the domU gets its time at startup, I''m
guessing it''s from the hypervisor and the hypervisor clock drifted away
from the dom0 one? Anyway, doing an ''hwclock --systohc'' in
dom0 and restarting the domU doesn''t change anything.
Ideally I''d want all clocks synchronized to the dom0 one, with NTPd
taking care of keeping dom0 synchronized.
If that matters, we''re planning an upgrade to Xen-4.1.2 but I
don''t know exactly when we''ll be able to roll that out; we
usually track upstream kernels for domU, and we''ll look at the
possibility to upgrade dom0 kernels to 3.1.x at a later time.
Any help would be appreciated, thanks.
--
Luca Lesinigo
