[Pool] time reset

Ryan Tucker rtucker at gmail.com
Fri Mar 25 02:42:29 UTC 2011


On Thu, Mar 24, 2011 at 03:38:32PM -0700, Chuck Swiger wrote:
> 1) Running ntpd in the Dom0/host ESX/host is very useful.  Keeping good time
> there means that good time will be available to all of the VMs/guests via
> independent_wallclock = 0, vmx option tools.syncTime = true, etc.

This is no longer the case with Xen, at least with the domU support in the
mainline Linux kernel.  The effectiveness and necessity of ntpd is
indistinguishable from real hardware, best as I can tell: if you don't run
ntpd, your clock will drift.

If you do run ntpd, everything is fine.  It has no problems keeping
accurate time, and the PLL stats show no differences from real hardware.
It makes for perfectly capable pool servers, too.

I am not 100% sure I fully understand the reasons why this is so; I nearly
got into a discussion about it on IRC a few days back, but had to leave
before I could look too much into it.  My working hypothesis is that
there's nothing magic between the domU kernel and the CPU, so clearly,
whatever hardware oscillator is driving the kernel's concept of the passage
of time does not suddenly become an ideal oscillator because of a
hypervisor.

(This is based on practical experience with Linode, Rackspace Cloud, and
Slicehost, all of whom include ntpd in their base deployment images.)  -rt

-- 
Ryan Tucker <rtucker at gmail.com>

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.ntp.org/pipermail/pool/attachments/20110324/f2e6666b/attachment.pgp>


More information about the pool mailing list