On 07/25/2013 01:28 AM, David G.Miller wrote:
> The problem of upgrading from FC-n to FC-n+1 is basically the same as 
> upgrading EL-n to EL-n+1. 

No; upgrading ELx to ELx+1 is like upgrading Fn to Fn+k(x), where k(x) 
is an element of an array of integer constants; x is the starting EL 
release, so k(3)=3 [RHEL3 was based on what I'm going to call 'Fedora 
Core 0,' which was the pre-fedora RHL 10 beta; see footnote 1]; k(4)=3, 
k(5)=6 (or 7, since some F13 packages showed up in EL6), and k(6) will 
probably be 7 or so.

Doing this without going stepwise through the Fedora releases is a 
challenge.  I forget how large of an increment preupgrade can do, but I 
remember doing it F12 to F13 to F14, and it had issues even going Fn to 
Fn+1, especially if any part of the massive yum transaction fails for 
any reason (it leaves the system with a half completed yum transaction 
that yum-complete-transaction simply won't deal with, and then you have 
to finish the upgrade manually and manually remove the older 
packages).... I have done this twice on two separate machines, one had 
issues going from F12 to F13 and the other one had issues going from F13 
to F14.  The Fn to Fn+1 upgrade path is somewhat expected to work; Fn to 
Fn+2 probably won't work correctly, especially if major changes are in 
both releases.

To get an EL5 system up to EL6 stepwise with yum, the only semi-sane way 
would be to upgrade EL5 (FC6 or so based) to F7, then to F8, then to F9, 
then to F10, then to F11, then to F12, and you might maybe possibly be 
able to get F12 up to EL6 reasonably easily. That's a k(5) of 6; EL6 to 
EL7 may be a bit harder, with a k(6) of 7 (assuming EL7 really does get 
based off of F19).

This of course goes through the KDE 3.5 to KDE 4.3 upgrade, so there's a 
very big difference if you're a KDE user.....

Having said that, back in the days when I was a user of Aurora SPARC 
Linux on some Sun Ultra hardware, yum upgrades were the norm, since 
install media wasn't spun for every release.  And it was not a single, 
simple yum transaction; certain packages had to be upgraded before 
others, and some had to be parachuted in with rpm.  And some things 
were, well, pretty difficult to deal with on a couple of the increments.

In the Ubuntu world, this is like taking Ubuntu LTS 6.06 straight to 
8.04, or worse.  I've done the 6.06 to 8.04 thing, by the way, and have 
no desire to repeat it.

In the Windows world.... well, there's actually several Youtube videos 
of that, and it's hilarious what had to be done to get a Windows 1.0 
install upgraded through every step to Windows 7 and 8. And the 
artifacts left behind..... [2][3]

Upgrades are hard to do properly, especially in a non-engineered system 
like a Linux distribution (I know, I know, upstream has engineers 
working on it and doing integration, but as long as there are packages 
whose own upstream is outside of TUV's control, it's not an engineered 
distribution in the strict sense.....).  (What I want is the 'Redneck' 
language back for installs........)

>  I wouldn't want to guarantee that an arbitrarily complex installation 
> will work though and the people who really want to upgrade are those 
> with really complex systems that they don't want to have to re-create 
> from a clean installation. 

Heh.  Try doing the upgrade with yum alone from EL4 to EL5 (or EL5 to 
EL6) on a server with an active PostgreSQL database.  For that matter, 
try using the unsupported 'upgradeany' anaconda boot argument and watch 
your working database go bye-bye.


[1] 
https://en.wikipedia.org/wiki/Red_Hat_Enterprise_Linux#Relationship_to_free_and_community_distributions
[2] https://www.youtube.com/watch?v=eY-UHdswFkg  (note the length! and 
goes to 8)
[3] 
http://rasteri.blogspot.com/2011/03/chain-of-fools-upgrading-through-every.html 
(note that the video is somewhat NSFW due to some text that is entered.....)