Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 10 Jan 2004 13:14:58 -0600 (CST)
From:      "Jon Noack" <noackjr@compgeek.com>
To:        "Ken Smith" <kensmith@cse.Buffalo.EDU>
Cc:        freebsd-current@freebsd.org
Subject:   Re: Look ma, the kernel don't compile no more...
Message-ID:  <3189.192.168.1.10.1073762098.squirrel@www.noacks.org>
In-Reply-To: <20040110181927.GA17659@electra.cse.Buffalo.EDU>
References:   <20040109121734.E78161-100000@moo.sysabend.org><87d69r4tsd.fsf@strauser.com> <20040110181927.GA17659@electra.cse.Buffalo.EDU>

next in thread | previous in thread | raw e-mail | index | archive | help
Ken Smith wrote:
> On Sat, Jan 10, 2004 at 11:06:58AM -0600, Kirk Strauser wrote:
>> Out of curiosity, why would multiple syncs be better than just one?
>> Wouldn't the final sync be the same whether you'd done several before =
it
>> or not?
>
> It all depends on where you're sync-ing from.  Keep in mind the
> cvsup mirror sites are exactly that - mirrors.  There is one
> central CVS repository.  If a developer is doing an update of the
> main CVS repository as several commits there can be several minutes'
> worth of time that what's in the CVS repository is "inconsistent"
> because what s/he committed first contains things that depend on
> things that s/he commits a little bit later (composing log messages
> can take time, etc.).
>
> cvsup-master itself mirrors from the main CVS repository roughly
> every 10 minutes I think.  So if it takes a snapshot between the
> developer's first and second commit and if you're using cvsup-master
> as where you sync from then doing two sync's 10 to 20 minutes apart
> can help with this sort of problem if you watch what gets updated
> to watch for stuff that "seems related".
>
> Virtually every other public cvsup mirror site will be mirroring
> from cvsup-master, but typically with about an hour between syncs.
>
> There are things that can make things a little bit more complicated
> than this.  For example last night's 5.2-RELEASE tag touches a huge
> number of files.  cvsup-master's load was around 50 for a while, and
> it took one machine I was watching over an hour to do the first sync
> after the tagging.  A mirror site won't make changes 'visible' to
> its downstream sites until it completes its own sync.

The second cvsup could get you the first portion of another developer's
update.  Likewise for the third, fourth, etc.  I think the most efficient
way is to cvsup once and build.  If the build fails, you can quite quickl=
y
check the commit logs (http://www.freshsource.org/commits.php is a great
resource for this) and re-cvsup if necessary.

If seems to me that multiple cvsups back-to-back are a drain on bandwidth
and server load with dubious benefits.

Jon Noack



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3189.192.168.1.10.1073762098.squirrel>