[time] Worried newbie - odd graphs?
Wed Apr 21 22:41:37 UTC 2010
I am a bit late to this thread. Just catching up on the list.
On Thu, Apr 01, 2010 at 15:18:50 +0000, Dave Hart wrote:
> You might want to save a
> snapshot of the offset graph as it is now somewhere
I saw a similar thing when I upgraded a Xen server from Debian Etch to
Before the upgrade, the Dom0 would run ntpd and the DomUs relied on Xen
keeping their clock close to the Dom0. With Lenny Xen and kernel, the Dom0
clock began to oscillate:
I eventually resolved it by setting independent_wallclock in all DomUs and
the Dom0 and running ntpd on each virtual machine. One VM with very low load
is the "master" and syncs to external stratum-1 sources while the other VMs
all peer with each other plus a few servers from the pool.
So far, ntp keeps the clock close to real time:
I intended to find out what changed between Xen and/or kernel versions and
how to get the Dom0 back in control of the Xen clock. But since ntpd is
small enough to put on all VMs, I never got around to dig into that.
> If my memory is correct, and you are running ntpd that was not built
> on the host running it, rebuilding from source (such as a download
> from http://archive.ntp.org/ntp4/ ) would presumably fix it by
> building a ntpd that matches your style of kernel loop.
IIRC, I didn't see any change with either pre-compiled packages or ntpd
built from source on the server itself.
My munin-plugin for the graphs is here:
Just symlink it to a variable name (see ntpq -c rv), e.g. for ntpd 4.2.4:
for t in offset frequency rootdelay stability stratum jitter noise; do ln -s ntpvar_ ntpvar_$t; done
for t in offset frequency rootdelay clk_wander stratum sys_jitter clk_jitter; do ln -s ntpvar_ ntpvar_$t; done
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 197 bytes
Desc: Digital signature
More information about the pool