what has 'yum update' done?

[lee]

Reindl Harald <h.reindl at thelounge.net> writes:

> Am 07.07.2013 19:21, schrieb lee:
>> Considering that Fedora does not have a working (and tested and
>> supported) upgrade method at all, such packaging guidelines seem very
>> questionable.  That means you're simply lucky if you manage to upgrade.
>> That users are quick to suggest to reinstall instead of trying to
>> upgrade seems to support this conclusion.
>
> i do not buy "you're simply lucky" after around 400 dist-upgrades
> with yum on workstations and production servers with all sort
> of services - i *never* rely on luck, i simply *test* and *prepare*
> upgrades after test them carefully on clones with note all needed steps

I don't have a clone of my computer around to test the upgrade process
on.  Do you have clones of all your workstations and servers which you
upgrade first, then replace the "original" ones with them and try them
out for six weeks to see if they work fine before upgrading the
"originals"?

>> So we have three options to upgrade:
>> 1.) the recommended fedup which probably doesn't work
>
> so no benefit compared to yum

Dunno, maybe it does work when I actually try it.

>> 2.) the untested and unsupported way using yum which might work or not
>
> the same for fedup

Yes, like I said, there is no working and tested upgrade method.

>> 3.) move away from Fedora and install a different distribution
>
> maybe your choice, mine is learn to understand every piece of my system

You can spend a lifetime with that and probably still won't because it's
just too much.  Do you know every bit about how the kernel works, just
to begin with?  Probably not ...  Do you fully understand emacs, with
all it's possibilities?  Probably not ...

Nowadays, I'm more concerned about actually using my computer for what I
want to do, learning along the way what I find I need to learn.
Updating the software shouldn't be an issue to worry about but a
straighforward process that works just fine.  It has never been an issue
with Debian.

It may be fun for you messing around with trying to upgrade Fedora.
Perhaps it doesn't matter to you when it doesn't work because you're
doing it on a clone anyway so it doesn't have any consequences.  It's not
fun for me.  When it doesn't work and leaves me with an unusable system,
I'm forced to go to the lengths of installing something else.  It's
a waste of my time.  Even when it does work, it's a PITA with Fedora
because as things are, I have to be prepared for it to not work, and I
have to disconnect the data disks before I even start.

>>> if you want simple fire&forget upgrades and not invest time
>>> and energy to learn how the system basically works and what
>>> major changes are meaning for you in case of upgrades use
>>> CentOS or whatever LTS distribution
>> 
>> That is a very different issue.  The problem is that there apparently
>> isn't a working way to upgrade Fedora
>
> there is - if you know what you are doing
>
> hence i managed dependency problems often enough due upgrades
> and that is why i use yum - you can fix problems before upgrade
> and even after upgrade before reboot

Well, I can only follow the upgrade instructions.  I did that last time
and was left with a mess.

Do you have some detailed documentation about how every bit of Fedora
works which I could read before trying to upgrade again?

>> Add to that that a simple 'yum update' made a severe mess, not only
>> basically putting the system into an unbootable state but also trying to
>> upgrade without even indicating that it would try this and without even
>> asking me if I wanted that.
>
> there doe sno upgrade happen magically

There wasn't any upgrade supposed to happen.  So do you think it a bug
with "yum update" that it "magically" tries to upgrade when it
shouldn't?

>> Only by examining the configuration of the boot manager I was able to
>> boot again.  That was just luck and I could have been left stranded with
>> a non-bootable system.
>
> which is all fixable
> been there, done that

See my other post, you can lose your data and/or be left with an
unusable system just by running "yum update" which you can expect to be
pretty harmless.

Imagine your car was like that ... like you fill the tank as usual and
suddenly it doesn't work anymore and you can't get to work or wherever
else you need to go.  Sure, it's fixable.  You check the tire pressure
and suddenly the brakes fail.  Sure, it's fixable --- if you didn't
crash into a tree.  So you get that fixed and change the oil, and the
engine falls out.  Sure, that's fixable, too --- if it didn't fall onto
your head.

Just like Fedora, it's fixable.  You can even get a new car, only you
need to start customising it to your needs again.  Half a year later,
you fill the tank and the engine stutters.  Wow cool, that's another
thing that's fixable.

But if you want your car for driving and to do what you have it for in
the first place and not for having something that you need to fix all
the time, maybe it's time to change the brand so you have a car that
actually works, don't you think?

>> The method for upgrading you suggest is not recommended and untested
>
> and you believe a fedup-upgrade 7 weeks after new updates are pushed
> out is tested - who has tested it? with what combination of installed
> pakcages which may make a big difference in possible conflicts

That's what I expect.  I don't know what they are actually doing, it's
very well possible that they just have their release dates and push out
whatever they have in whatever state it happens to be at that time.  If
they do that, they shouldn't call it a release and instead make it clear
to the users that what they get is yet another alpha stage of an
untested Linux distribution that happens to be called Fedora.

But they don't do that.  They tell users to use fedup to upgrade.  Now
look at [1] and then tell me what they really expect and if you are sure
that you can expect from users to have clones of their computers around
to test update processes and that every user understands in every bit of
detail how Fedora works.  And please, point me to a documentation that
explains in every bit of detail how Fedora works so that the users can
successfully upgrade.


[1]: http://fedoraproject.org/wiki/User_base

>> It might be a great a way to do it or it might fail.  That doesn't have
>> anything to do with learning "how the system basically works and what
>> major changes are meaning for" me.
>
> surely, if you know your system well you are able to verify package
> consistency and configuration *before* reboot and fix things
> if needed after yum-upgrade

Then what?

>> Atm, I'm undecided if I even should try to upgrade or save myself the
>> hassle and just move on to another distribution which might be more
>> reliable. 
>
> that has nothing to do with "reliable"
> Fedora *is* relieable
> if not i would not run more than 20 production servers
> with it since many years

Sure it's about reliability.  Running "yum update" tricking the
unsuspecting and unprepared user into upgrading and putting them at the
risk of losing their data and being stranded with a non-working system
constitutes unreliability.  Their recommended upgrade procedure
consistently failing miserably constitutes unreliability.  The missing
of a working and tested upgrade procedure altogether constitutes
unreliability.

If I had known that it's so troublesome to upgrade, I wouldn't even have
tried Fedora.  Now I would recommend to stay away from it.

Fedora on production servers?  Sure, if you're prepared to take them out
of production for however long it takes to fix problems an upgrade ---
and now even a simple update --- may cause, or to re-install them,
you're fine.  I don't want such unreliability and unpredictability, but
that's just me.

>> I don't want to have problems with updating all the time, and
>> so far, Fedora promises to continue to give me these problems
>
> maybe it is not the right distribution *for you*

Maybe --- I'd have to see how upgrading goes this time.  After the last
time, I expect it to be fixed and working now.  They do have the concept
of release critical bugs, or don't they?

> if all distributions would be perfect for everyone why would we
> have more than one?

Perhaps it is part of a process to find a perfect one?


-- 
"Object-oriented programming languages aren't completely convinced that
you should be allowed to do anything with functions."
http://www.joelonsoftware.com/items/2006/08/01.html