From: Matthias Kilian (kilioutback.escape.de)
Date: Thu Sep 02 2004 - 17:21:53 CDT

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On Tue, Aug 31, 2004 at 09:56:31PM +0200, Hannah Schroeter wrote:
> >If client and server are 5 minutes off from each other, so what?
>
> >If the NFS implementation isn't completely broken, editing and building
> >on the same client shouldn't be affected by time glitches between client
> >and server.
>
> How so? Does the client stub cache the times? And if so, what happens
> when the cache entries get evicted/replaced? Or does the client set
> *its* times on the file?

After sleeping, thinking, reading, I come to the conclusion that I
probably wrote complete bullshit. Sorry for that.

May be I memoized some ill-caching client I used several years ago.

> >BTW: I recently got a gmake warning about future modification times on a
> >gentoo linux box synchronized to some time hopping (i.e. bad configured)
> >ntp servers. No NFS involved at all. IMO, that should be impossible.
>
> That should depend on the ntpd's options, i.e. whether it steps the
> time at all, or whether it uses only adjtime.

No special configuration, just two server entries. May be a bug in ntpd
or in adjtime.

For openntp, I didn't yet test what happens if it's servers are hopping,
but hopping servers are (or should be) a pathologic case anyway.

Ciao,
        Kili

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]