Date: Mon, 19 Dec 2005 21:58:40 -0500 From: Bill Vermillion <bv@wjv.com> To: "Matthew D. Fuller" <fullermd@over-yonder.net> Cc: freebsd-isp@freebsd.org Subject: Re: upgrades Message-ID: <20051220025840.GA17224@wjv.com> In-Reply-To: <20051220024711.GP63497@over-yonder.net> References: <Pine.BSI.4.05L.10512191705310.3604-100000@mail.lanline.com> <20051220002815.GB15974@wjv.com> <20051220024711.GP63497@over-yonder.net>
next in thread | previous in thread | raw e-mail | index | archive | help
I know you'll find this hard to believe, but on Mon, Dec 19, 2005 at 20:47 , Matthew D. Fuller actually admitted to saying: > On Mon, Dec 19, 2005 at 07:28:15PM -0500 I heard the voice of > Bill Vermillion, and lo! it spake thus: > > As to having one CVSsup on one machine and using nfs for the others > > that is up to you. > I did this for years. Saved bandwidth (and load on the cvsup > servers), saved disk space, saved processor, and saved me headaches > trying to keep track of what built when. If you've got a bunch of > systems that are otherwise fairly identical anyway, I'd recommend it. All mine do the CVSup in the wee hours of the morning and I'm on Level 3 backbone - and we haven't hit our bandwidth cap yet - even with one server that is always #1 on Google and MS. > > Where you can run into stale binaries are things from ports where > > things change- and if you forget to perform portupgrade and just do > > a new install. Sometimes locations change. > Ironically enough, I just last month or so had a major problem with a > stale binary. I tried for a week to upgrade something gnome-related > (libgnomeprint, I think?), and it kept bombing out in the build with > really weird errors. I finally tracked it down to its dependancy on > bison, which port wasn't even installed. And it didn't install the > port because it found /usr/bin/bison (datestamp Dec 30, 1999) and used > it. So, yes, doing a `ls -lt` every once in a while is a good > thing... Ah - that rings a bell - an out of tune bell :-). But in my case slightly inverted. I had soemthing that needed to update Bison and it kept failing with gmake ?? which was called in the upgrade process. I started saving saving logs and found when it dropped into a sub-directory it lost the global varialbes it was expecting [or something similar to that]. I had failures on 5 different servers. And I went to the 6th server - a newer one - and things just flew right along. This was about 2 weeks after trying to find the problems on the other servers. It turns out that while I was using the KSH - the real one from David Korn - the machine that worked had a newer version of ksh93. And update on the other machines fixed that one. A new Bison built and everything was fine. Unix systems are the greatest adventure game the world has seen. Bill -- Bill Vermillion - bv @ wjv . com
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20051220025840.GA17224>