On Mon, 16 Mar 2015 09:46, Per Jessen wrote:
Per Jessen wrote:
ianseeks wrote:
On Monday 16 Mar 2015 07:45:21 Per Jessen wrote:
On Sunday 15 Mar 2015 00:16:49 Per Jessen wrote:
On Sat, 14 Mar 2015 14:54:44 +0100 Andrei Borzenkov wrote: > Per Jessen wrote: >> On Sat, 14 Mar 2015 11:36:02 +0100 Andrei Borzenkov wrote: >>> Per Jessen wrote: [snip] Well, it's quite clearly systemd, but as Andrei says, systemd is just
ianseeks wrote: the messenger. Looking at a xen guest this morning, I see the message twice every minute, steady as clockwork :-) In total 1439 messages right now.
so is it just systemd monitoring the change on the system and reporting it rather than actually changing it?
I think so yes. Monitoring or being notified by <something> that is monitoring. What I'd like to understand is -
systemd appears to be monitoring it:
from src/core/manager.c:
/* Uses TFD_TIMER_CANCEL_ON_SET to get notifications whenever * CLOCK_REALTIME makes a jump relative to CLOCK_MONOTONIC */
I won't pretend to understand what that means :-(
A hunch: the 'internal' clock in the xen guest runs to fast/slow, compared to RTC and is reset/corrected by ntp once a minute. Thus the 'MONOTONIC' is broken (a time stamp occures twice), and the message is send. Check the mobo nvclock / kernel time div (adjtimex, ntp logs), and do the same in the xen client. At more than a 1/1000 (1ms per 1s) systematic drift / divergence it gets ugly (in terms of messages). The better the systematic drift is accounted for, the less ntp has to do, the less messages you get. - Yamaban. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org