Re: Time sync to Windows. How?
Ron Johnson said...
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> marc wrote:
> > Joshua J. Kugler said...
> >> On Friday 14 July 2006 13:07, marc wrote:
> >>> Hi,
> >>> I've managed to get Linux through the door at a client's site
> >>> <whoop, applause> However, it is necessary to time sync
> >>> desktop client's to a Windows server. (This is a strict
> >>> requirement for reasons that I won't go into. Yes, I know how
> >>> trivial it would be with a Linux server, etc.)
> >>> Windows has a "service" called Windows Time, but I haven't
> >>> managed to sync to it from the Linux boxes. Is it possible?
> >>> In an attempt to approach this in a sane manner, I installed
> >>> NetTIme on a Windows box for testing. This worked, up to a
> >>> point - the Linux boxes could sync to it - but it has a huge
> >>> flaw in that it refuses to act as a time server when it can't
> >>> access the sources it syncs from - who designs this stuff?
> >>> Since the server s not always connected to the net, this
> >>> means that the clients lose sync, which is not acceptable in
> >>> this case.
> >>> Basically, all I need is a time service on Windows that will
> >>> sync when sources are available, but continue to allow
> >>> clients to sync to it when they are not.
> >>> Any ideas or suggestions?
> >> Should get you going:
> >> http://www.google.com/search?q=ntp+windows
> > I'm a little beyond the 101 stage, thanks. I'm asking for
> > suggestions for a solution from folk who likely have experience
> > with the problem.
> > I've already googled, read, installed, etc. There's little time
> > left to provide a solution before we'll resort to a Windows-only
> > setup and the chance for Linux to be deployed in a Windows-only
> > shop will have passed.
> > I've yet to find an time server for Windows that will continue to
> > server clients when its external sync is absent.
> Ah, so you've got a *Windows* question. Why are you asking us?
Because there are Linux clients and it's likely that someone else will
have tackled the same problem before.
> Still, from Googling, I found this:
> W32Time is based on the Simple Network Time Protocol (SNTP)
> as specified in RFC RFC 1769 (now superceded by RFC 2030).
> SNTP is designed to ensure *loose* synchronization only,
> which in the W32Time implementation means the clocks of all
> Windows 2000/XP/2003 machines in a forest will agree within
> 20 seconds of one another (or 2 seconds difference within a
> particular site).
> Sounds pretty bad to me.
That's why you strip SNTP and apply NTP. And, in any case, you haven't
asked about the degree of synchronisation that is required, so "loose"
might be sufficient.