Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 26 Feb 2006 23:33:27 -0600
From:      "illoai@gmail.com" <illoai@gmail.com>
To:        "pobox@verysmall.org" <pobox@verysmall.org>, freebsd-questions@freebsd.org
Subject:   Re: advice on upgrading production FreeBSD 5.4
Message-ID:  <d7195cff0602262133o712f52b5qae8fce16edf8dabe@mail.gmail.com>
In-Reply-To: <cb5206420602261419y519ea8e5g972fe3fc15a04488@mail.gmail.com>
References:  <4401F701.10202@verysmall.org> <cb5206420602261419y519ea8e5g972fe3fc15a04488@mail.gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On 2/26/06, Andrew Pantyukhin <infofarmer@gmail.com> wrote:
> On 2/26/06, pobox@verysmall.org <pobox@verysmall.org> wrote:
> > I inherited a production FreeBSD 5.4 used as a web/mail server (Apache,
> > PostgreSQL, php, qmail, vpopmail, Courier). Could anybody help me with
> > information about a web resource on upgrading such system (all, OS only
> > or component by component) with minimum downtime. I looked into cvsup
> > and portupgrade - is this the right way for production systems or is
> > there another one? For example is it possible to have the older version
> > running until the new one downloads/compiles and then to replace it
> > within seconds? Also - what if the new version does not work correctly =
-
> > is it possible to keep the old and revert to it.
>
> First, you'd better upgrade to 6.1. Read /usr/src/UPDATING
> and handbook for that.
>

I'm not sure if I had a working machine that was up on its security patches
that I would upgrade, where admittedly 5.4 to 6.x is not a huge jump
(compared to 4 -> 5 or even 3 -> 4).  Assuming all goes well your down-
time should be the time it takes the server to reboot, but if (very big)
your buildworld has some hidden fault (personal experience when /bin/sh
would dump core on every invocation, very difficult to fix) you could be
looking at a couple of days of downtime.
Whereas I believe that the 6.x series is much better than 5.x, I'm not
convinced that the risk/reward payoff is that great, doubly so given
that it's a production machine.  5.4, while obsolete from a numerical
standpoint, will be useable for a long while: years, probably.

If the hardware itself is nothing special, or replaceable for less than the
cost of downtime, you might look into putting up a second server
running the new software, test carefully for a week or so, and then
gracefully transition.

All of this is speculation.

As a postscript: the tales of 5.x to 6.x upgrades have been for the most
part very painless, but in production systems conservatism leads to
happy customers.

--
--



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?d7195cff0602262133o712f52b5qae8fce16edf8dabe>