[BUGS] question on up keeping of a production server

[Andrew Reilly]

On Thu, Jun 05, 2008 at 09:25:46PM +1000, Brad Rushworth wrote:
> How easy is this:
> 
> 
> 
> [root at 220 ~]# ./freebsd-update.sh fetch
> [root at 220 ~]# ./freebsd-update.sh install
> Installing updates... done.

Don't you have to reboot after that?  Is there an equivalent to
mergemaster?

I've always updated from source[*], tracking -STABLE approximately
weekly on my main workstation.  I track STABLE sporadically,
when something interesting or important has happened, on
the server system I manage.  On my workstation, the csup
happens on a cron job, and I run a four-line script that runs
buildworld, kernel, installworld, mergemaster manually.  Against
recommendations, I don't shut down to single user to install,
and only reboot once.  I probably only have two minutes downtime
(for the reboot cycle) per week, this way, and only a couple of
minutes of eyeball time for the mergemaster script.  I do check
UPDATING first though, to make sure that there isn't any manual
heavy lifting to do.

I've been toying with the notion of firing off the buildworld,
kernel steps on the cron job, but have so far been
insufficiently bothered by the current method to get around to
it...

[*] except for two situations: new hardware usually starts with
a basic ISO image CD install and then a csup to -STABLE, and
so did the major version step when UFS2 and softupdates was
introduced: I reformatted my drives for that one.  Back in the
days of POTS modems I would do hardware upgrades by migrating
disks between machines, to avoid the downloads, but that's too
much to bother with, now.

Cheers,

-- 
Andrew