Re: Yum and Fedora 16 -- focus on "new/moved filesystems"





Am 10.02.2012 14:43, schrieb Timothy Murphy:
Reindl Harald wrote:

sorry, but permanently reinstall the OS is a windows-thing
and especially unnaceptable if all 6 months a new version
is available and you have to maintain 2, 5, 10, 20 machines

I don't agree - at least if one has 2 or 5 machines to deal with.
I always do a fresh install on a new partition,
leaving the /home partition untouched.
(I don't think this is a "Windows thing";
I don't even know if it is possible under Windows.)

you are a simple desktop-user with your stuff
in /home, if you are develop software and
integrate workflows in a system for managment
and so on /home does not interest you much

One reason is that in my experience there is a non-zero probability
that the new version will not work, perhaps because of a driver problem.
In fact I would guess that over all the Fedora systems I have installed
there has been a 25% initial failure@machine rate,
usually to do with video card drivers.

a scripted dist-upgrade for 20 servers with yum takes around
two hours (proveable by logs) inlcuding download from local
repo-cache

What is the script?
It seems to me it would have to be pretty complicated.
I certainly wouldn't trust any script I wrote to do this.

* prepare the dist-upgrade
* test it with snapshots
* rebuild weak packages (missing unit-files, dependencie problems..)
* build up an internal repo
* have all machines ONLY this repo as source
* have all machines cloned from the same master
* have all machines exepct the repo-cache never seen external repos

after that you can test a dist-upgrade until your finger bleeds
and you are knowing EXACTLY what will happen on the live-machine

you can prepare even configurations for services before
rollout - again you know exatly what happens, what have
to be prepared and what excatly commands you have to type
after the dist-upgrade before the reboot and put
them in a script on each target machine
_________________________

finally a simple shell scipt
"ssh root@taget-machine1 'yum -y commands; /finish-script.sh"
"ssh root@taget-machine2 'yum -y commands; /finish-script.sh"
"ssh root@taget-machine 3'yum -y commands; /finish-script.sh"
_________________________

i have done my homework in 2008 and planned a working infrastructure
where new servers are cloned from the same master and i did EVERY
dist-upgrade (F10, F11, F12, F13, F14) this way without a single
problem because i know what i am doing and holding my machines clean

and that is why my understanding breaking this in one or
the other way is simply not existent and called an unacceptable
step backwards, and yes i have a VOIP-Server which has seen every
single update since FC7 until F15 via yum because as said i
know what i am doing and i take the time to prepare things
well BEFORE they are needed

that is also the reason for my non existing understanding
for put dumb service restarts on yum update in every package
instead a global 0/1 setting forxing me tu rebuild each
damned server-package and bringing no benefit at all because
it is not well thought believing in better security doing
this automaticallyl in the case of security updates not realizing
that this would as example not help if a php-security update
arrives which does not restart httpd, or a gd-library update
fixing security bugs and affecting all sorts of other packages
including php and at least apache

such things are low-brained decisions without looking
at the big picture

Attachment: signature.asc
Description: OpenPGP digital signature

--
users mailing list
users@xxxxxxxxxxxxxxxxxxxxxxx
To unsubscribe or change subscription options:
https://admin.fedoraproject.org/mailman/listinfo/users
Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines
Have a question? Ask away: http://ask.fedoraproject.org