Date: Fri, 14 Sep 2001 12:28:57 -0700 (PDT) From: John Polstra <jdp@polstra.com> To: hubs@freebsd.org Cc: marcel@xcllnt.net Subject: Re: Randomizing access to CVSup mirrors shows inconsistency Message-ID: <200109141928.f8EJSv169439@vashon.polstra.com> In-Reply-To: <20010914120655.A1150@dhcp01.pn.xcllnt.net> References: <20010914120655.A1150@dhcp01.pn.xcllnt.net>
next in thread | previous in thread | raw e-mail | index | archive | help
In article <20010914120655.A1150@dhcp01.pn.xcllnt.net>,
Marcel Moolenaar <marcel@xcllnt.net> wrote:
>
> I recently changed my setup to not hardcode a single CVSup mirror,
> but instead define a group of mirrors and randomize the access.
> The egocentic intend here is to be resistent to mirror failures,
> but I can't deny that it helps distribute the load on the mirrors
> as a nice side effect :-)
>
> Unfortunately, it doesn't work without glitches. The biggest and
> most annoying glitch is that I see my complete repository touched
> by SetAttrs. Not always, but depending on the random order in
> which I use the mirrors. I noticed it when I used cvsup7.FreeBSD.org
> after using cvsup3.FreeBSD.org.
> Using cvsup13.FreeBSD.org after cvsup7.FreeBSD.org exhibits another
> problem. I see under CVSROOT that commits have been made, but the
> collections aren't updated. After aborting cvsup13, I restarted
> the script and got cvsup14. This touched my complete repository
> again.
If you are getting a bunch of SetAttrs then something is wrong. I
pull my updates from many random mirrors and I never see this
problem.
It could be that one or more of the mirrors is still running a server
with the s1g bug. See
http://people.freebsd.org/~jdp/s1g/
for instructions on how to find out which version a mirror is
running. If you yourself have not upgraded your CVSup client,
you would also see this problem.
These SetAttrs could also be caused by use of an inconsistent umask
setting when your script runs, or by cron jobs you have forgotten
about. All of the possibilities are discussed in the CVSup FAQ at
http://www.polstra.com/projects/freeware/CVSup/
> Ok, now the questions:
> 1. Is it actually valid to randomize access to mirrors?
Yes, as long as you realize that moving quickly from one mirror to
another may actually regress your files, because the mirrors are not
perfectly in sync with each other.
> 2. Can it be something in my configuration (attached)?
Could be. See the URLs above.
> 3. If it has something to do with the version of cvsupd,
> is it possible to automaticly regulate access to
> mirrors based on version information? (a simple
> abort or proceed will do just fine)
It shouldn't be caused by the server version, except for the s1g
bug. Most of the mirrors are upgraded now. If you find any that
aren't, please write to their maintainers and ask them to upgrade or
shut down the service. Maintainers are listed in the handbook along
with the list of mirrors.
John
--
John Polstra jdp@polstra.com
John D. Polstra & Co., Inc. Seattle, Washington USA
"Disappointment is a good sign of basic intelligence." -- Chögyam Trungpa
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hubs" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200109141928.f8EJSv169439>