You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@subversion.apache.org by Andy Levy <an...@gmail.com> on 2007/01/04 13:46:29 UTC

Importance of clock synchronization?

For reasons I don't fully understand, the system clock on our VMWare
ESX server, which will soon host a virtual machine in which we'll run
our new Subversion server, is not synced with the rest of the world -
it's about 9 minutes behind. As a result, any virtual machine running
on this host is also behind.

All our other servers, and desktops, are synced with a central server,
which syncs with an external NTP server.

I've asked our server admin to get the ESX server straightened out,
but if that doesn't happen before we go live with the new Subversion
server, what's our risk level? Is this time difference going to cause
trouble to the point that we can't trust the repository 100%?

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org

Re: Importance of clock synchronization?

Posted by Erik Huelsmann <eh...@gmail.com>.
On 1/4/07, Andy Levy <an...@gmail.com> wrote:
> On 1/4/07, Erik Huelsmann <eh...@gmail.com> wrote:
> > On 1/4/07, Andy Levy <an...@gmail.com> wrote:
> > > For reasons I don't fully understand, the system clock on our VMWare
> > > ESX server, which will soon host a virtual machine in which we'll run
> > > our new Subversion server, is not synced with the rest of the world -
> > > it's about 9 minutes behind. As a result, any virtual machine running
> > > on this host is also behind.
> >
> > Only if syncing it later means a jump back in time, because then you
> > may get non-continuously-increasing commit times which is technically
> > no problem, but will break -r {date} revision specification.
> >
> > But, since your clock is behind, this is not a problem.
>
> Great news, thanks.
>
> One followup though. If the clock was ahead by 10 minutes instead of
> behind, couldn't I just disallow commits (shut down Apache, the whole
> box, or just ask people to stay out) for 15 minutes and then resume as
> normal?

That would totally solve the issue I raised, yes.

> More a curiosity than anything else, since we seem to be in the better
> of the 2 out-of-sync situations by being behind.

bye,

Erik.

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org

Re: Importance of clock synchronization?

Posted by Andy Levy <an...@gmail.com>.
On 1/4/07, Erik Huelsmann <eh...@gmail.com> wrote:
> On 1/4/07, Andy Levy <an...@gmail.com> wrote:
> > For reasons I don't fully understand, the system clock on our VMWare
> > ESX server, which will soon host a virtual machine in which we'll run
> > our new Subversion server, is not synced with the rest of the world -
> > it's about 9 minutes behind. As a result, any virtual machine running
> > on this host is also behind.
> >
> > All our other servers, and desktops, are synced with a central server,
> > which syncs with an external NTP server.
> >
> > I've asked our server admin to get the ESX server straightened out,
> > but if that doesn't happen before we go live with the new Subversion
> > server, what's our risk level? Is this time difference going to cause
> > trouble to the point that we can't trust the repository 100%?
>
> Only if syncing it later means a jump back in time, because then you
> may get non-continuously-increasing commit times which is technically
> no problem, but will break -r {date} revision specification.
>
> But, since your clock is behind, this is not a problem.

Great news, thanks.

One followup though. If the clock was ahead by 10 minutes instead of
behind, couldn't I just disallow commits (shut down Apache, the whole
box, or just ask people to stay out) for 15 minutes and then resume as
normal?

More a curiosity than anything else, since we seem to be in the better
of the 2 out-of-sync situations by being behind.

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org

Re: Importance of clock synchronization?

Posted by Erik Huelsmann <eh...@gmail.com>.
On 1/4/07, Andy Levy <an...@gmail.com> wrote:
> For reasons I don't fully understand, the system clock on our VMWare
> ESX server, which will soon host a virtual machine in which we'll run
> our new Subversion server, is not synced with the rest of the world -
> it's about 9 minutes behind. As a result, any virtual machine running
> on this host is also behind.
>
> All our other servers, and desktops, are synced with a central server,
> which syncs with an external NTP server.
>
> I've asked our server admin to get the ESX server straightened out,
> but if that doesn't happen before we go live with the new Subversion
> server, what's our risk level? Is this time difference going to cause
> trouble to the point that we can't trust the repository 100%?

Only if syncing it later means a jump back in time, because then you
may get non-continuously-increasing commit times which is technically
no problem, but will break -r {date} revision specification.

But, since your clock is behind, this is not a problem.

HTH,

Erik.

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org