kashani wrote:
> I am curious about how other people do updates once
they've decided
> there is no issue and how hardware/load plays a part in
the process.
Generally I am more concerned with a stable, predictable
build process than
compile times. What wastes more time than anything else is
getting 10
hours into the build of a server and hit a brick wall that
takes 4 times
the amount of time to work around. An example, yesterday I
was building a
hardened installation from stage1. Generated a stage3, got
all of my
binary packages built, installed in a vm to test. Apache
would not start,
libarutil.so was looking for gdbm_errno, which appears to be
missing. I
did not have gdbm in my use flags so I was scratching my
head. Ran
portageq envvar USE in my stage1 and stage3, for some reason
gdbm is
getting included in the use flags in the stage3. Exact same
make.conf,
exact same use flags, exact same profile, different results.
I can't
figure out why it is happening, my best guess is that
something is
poisoning the environment from the host server in my stage3
chroot. I
changed my use flags to USE="-* ....." and
started back from scratch.
Another wasted day.
The moral to my story is that compile times are a pointless
benchmark if the
process used to build the packages is corrupt or
inconsistent in some way.
Get the process nailed down and you can completely automate
your compiles,
start them at 5:00 pm as you head out the door ;)
To answer your question ;) At work I have one server that
does all of the
compiling, a dog slow dual 600Mhz with 2G of ram. At home,
I use my mythtv
pvr ;) It is a P4 3Ghz with 2G ram, I often watch a
recording while
recording another one, while building in the background
(unless compiling
qt or kde packages with kdeenablefinal, will produces burps
in video). It
tends to take about 1/5th of the time as a dual 600Mhz ;)
--
gentoo-server gentoo.org mailing list
|