Cross-grading from i686 to x86_64: it is possible (but unsupported)

2012-04-21 Thread Roberto Ragusa
Hi,

after reading a lot of advice about not doing it, and only a couple
of reports from guys who did it, I finally decided to try to
cross-grade my Fedora installation from i686 to x86_64.

Everyone says you should just reinstall; they probably do not
value their customizations. My system has been upgraded from Fedora
Core 3, and has been moved on a multitude of different hardware.
Preserving /home, /usr/local, /opt, /etc is still not enough,
you then have to be sure everything you had installed is there
again: big job (maybe you do not have the installation package
anymore).

I can't describe all the details for the cross-grading, but I want
to try to be helpful for others wanting to try that.

The transition was a F14-i686 to F14-x86_64 (end-of-life!).
On a second step there will be an upgrade to F15 and F16.

The system had many rpms from additional repos (fusion, atrpms,
java-sun from jpackage, a few rpms from some vendors).

The system was running the latest x86_64 (!) kernel for F16 (!!) .

For the record: running an x86_64 kernel instead of the i686.PAE
works perfectly, and it is better for speed and stability (why Fedora
did not push this feature I do not know). I also discovered that
running a F16 kernel on F14 is OK; yum complains about some dependencies
(module-init-tools, IIRC), but it solved some hibernation issues.

FIRST STEP: TOTAL FILE-LEVEL BACKUP OF THE SYSTEM (rsync to another machine;
this is a habit for me)

I've actually done the migration by yum in a graphical shell. (!)
So we already have collected a good quantity of "no, do not do that"
things. :-)

The most difficult part came now.
I wanted to start installing x86_64 stuff in parallel to i686 and
it was not exactly clear to me on what base rpm and yum decide the arch.
The final trick was, IIRC, just install rpm.x86_64; it appears to
contain the "info" about the running architecture. Some things
were required to be installed at the same time (mostly glibc).
I had a complete local mirror of F14 packages to avoid going to the network.

The "file `which rpm`" command started to say that it is an ELF64,
so I moved to yum. This is more complicated because it takes some
Python dependencies with it.
I repeated tried "rpm -ihv --test" and added more and more dependencies
on that command until it succeeded. Then "yum clean all".

In this phase I discovered that if you have a 64bit rpm and a 32 bit yum
the rpm database gives you strange errors. Every time you switch
from one tool to another the database appears to be corrupted, but it
isn't. The solution is to just remove a few tmp files: see RH bugzilla 553998.

At this point I started to use yum to install x86_64 stuff.
Started with small things (grep, bash), then moved to things with more
dependencies and then started to involve heavy things (for example, evince
will drag some GNOME, and so on). As I was running KDE, I used the GNOME
desktop as a guinea pig. :-)
Having both i686 and x86_64 version fails in some cases because the shared
files are not identical. I had to remove the i686 version in some cases and
just install the x86_64 one. Some "rpm -e --test" and "rpm -e --nodeps" were
necessary to understand how big was the risk of breaking the system at every
step.

A really big issue: i686 rpms from atrpms which I was not able to move to x86_64
because the f14-atrpms repository has been deleted from the net. Aaaargh.
Only a couple of stale mirrors were still there, I catched everything from them,
and also had to downgrade versions or remove some stuff and take a note of
reinstalling it later (maybe on F15). Why doesn't atrpms have an archive
repository like fedora???

For safety, I installed the rpms for KDE by logging in via ssh. The kde 
environment
was still running and it was unaffected by the process, anyway.

Some bash tricks involving "rpm -qa", "grep" and "yum install" took care of
installing as much as possible as x86_64, removing i686 when necessary to 
resolve
(apparent) conflicts.
It is fun to see there are -devel packages for both archs.
At the end I had an almost completely biarch system.

So reboot. (No change to grub, the kernel is still f16 x86_64).
The boot goes well.
X doesn't start: no driver. I had a /usr/lib path in /etc/X11/xorg.conf, which 
I had
to replace with /usr/lib64. Solved.
But KDE doesn't start. Hmmm. GNOME does, XFCE does, fluxbox does.
After some deep investigation, the solution was to just remove the content of
/var/tmp/kdecache-*, which is probably saved in an architecture dependent 
format.

As a cleanup, I started to remove i686 stuff. My idea is to only keep some
libs for compatibility with i686 only applications.
Started doing some of the work manually (yum remove and be sure you are only 
removing
stuff you have on x86_64 too).
Then wrote more bash one-liners: list all the rpms, find everything i686 which 
is not
dependent on anything else and exists in x86_64 format too; remove the i?86.
Repeat it a few times to escalate deeper in depend

Re: ifcfg-ethX voodoo for a second IP on the same interface

2012-04-21 Thread Sam Varshavchik

Ed Greshko writes:


On 04/21/2012 01:41 PM, Ed Greshko wrote:
> You're right about the doc and NetworkManager... I had to set it to no
> > and bring down the service to make it work...  But then I'm using F17
> > beta so maybe NM is broken???

No, AFAIK, NM won't handle that type of configuration.

You either do it the way you did, without NM.  Or, you do it the way I did,  
with NM.


The other "downside" is that the NM GUI (at least on KDE) doesn't allow for  
adding a

second IP address.  Needs to be done manually.


Yeah, I tried having an alias file every which way imaginable, and just  
couldn't get it to work while having NetworkManager and DHCP run things.


NM, and DHCP, are very convenient, even if, for all practical purposes, the  
assigned IP addresses are static. That does make things easier. Looks like  
I'll just go with executing "ip addr add" manually, when I need to bring up  
a second IP address. It's not really the end of the world, but it would've  
been nicer if it were automated. I think that NM has a post-* script  
somewhere, that gets executed when it brings up an interface with a DHCP- 
acquired IP address, that I can stick that into…




pgpZJ9DKrvni4.pgp
Description: PGP signature
-- 
users mailing list
users@lists.fedoraproject.org
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


Re: nouveau tv-out

2012-04-21 Thread Konstantin Svist

On 04/19/2012 08:40 PM, Sam Varshavchik wrote:

Konstantin Svist writes:

I tried connecting a laptop s-video out to a tv last night but 
couldn't see a software option for it anywhere.


Most laptops have a key combination to activate external video.

Running Fedora 16 x64 XFCE with nouveau driver. Checked xrandr and F 
-> Settings -> Display applet.


Also tried a VGA to s-video adapter, same story.

Tried installing kmod-nvidia - X refuses to start altogether.

What should I try next?


Examine your keyboard. Look for an icon on some key that looks like a 
monitor. The icon is usually painted with the same color as some other 
key that's typically labeled "Fn". That's your hint. There may be more 
than one key, to cut over your video completely, to the external port, 
or to have it go both to the laptop and to the external video.




Pressing Fn+[CRT/LCD] button does nothing, either.


--
users mailing list
users@lists.fedoraproject.org
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


syslogd in F-16 -

2012-04-21 Thread Bob Goodwin - Zuni, Virginia, USA


   DD-wrt has a logging function which I am unable to make work
   per: http://www.dd-wrt.com/wiki/index.php/Logging_with_DD-WRT
   since it apparently needs syslogd which has been replaced with
   syslog-ng if I understand what I've found.

   Is there a way to deal with this?

   Bob

   -- 
   http://www.qrz.com/db/W2BOD


   box7

--
users mailing list
users@lists.fedoraproject.org
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


Re: syslogd in F-16 -

2012-04-21 Thread Todd Zullinger

Bob Goodwin - Zuni, Virginia, USA wrote:
DD-wrt has a logging function which I am unable to make work per: 
http://www.dd-wrt.com/wiki/index.php/Logging_with_DD-WRT since it 
apparently needs syslogd which has been replaced with syslog-ng if I 
understand what I've found.


That page is referring to syslogd on the dd-wrt system, as far as I 
can tell.  It mentions in passing that if you want to send logs to a 
remote host, that the remote host needs to be running a syslog daemon, 
but makes no mention of what daemon that should be.  Any syslog daemon 
on the remote host would suffice.  I have rsyslog at home with dd-wrt 
logging there.


--
ToddOpenPGP -> KeyID: 0xBEAF0CE3 | URL: www.pobox.com/~tmz/pgp
~~
Genius is 1% inspiration and 99% perspiration, which is why engineers
sometimes smell really bad.
-- Demotivators (www.despair.com)



pgp7g2LQdgSXd.pgp
Description: PGP signature
-- 
users mailing list
users@lists.fedoraproject.org
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


Re: syslogd in F-16 -

2012-04-21 Thread Bob Goodwin - Zuni, Virginia, USA

On 04/21/2012 01:19 PM, Todd Zullinger wrote:

Bob Goodwin - Zuni, Virginia, USA wrote:
DD-wrt has a logging function which I am unable to make work per: 
http://www.dd-wrt.com/wiki/index.php/Logging_with_DD-WRT since it 
apparently needs syslogd which has been replaced with syslog-ng if I 
understand what I've found.


That page is referring to syslogd on the dd-wrt system, as far as I 
can tell.  It mentions in passing that if you want to send logs to a 
remote host, that the remote host needs to be running a syslog daemon, 
but makes no mention of what daemon that should be.  Any syslog daemon 
on the remote host would suffice.  I have rsyslog at home with dd-wrt 
logging there.






   Ok, I guess I didn't understand everything I read. It looks like
   I have rsyslog running.

   [bobg@box7 ~]$ service rsyslog status
   Redirecting to /bin/systemctl  status rsyslog.service
   rsyslog.service - System Logging Service
  Loaded: loaded (/lib/systemd/system/rsyslog.service;
   enabled)
  Active: active (running) since Sat, 21 Apr 2012
   09:11:45 -0400; 4h 17min ago
 Process: 865 ExecStartPre=/bin/systemctl stop
   systemd-kmsg-syslogd.service (code=exited, status=0/SUCCESS)
Main PID: 908 (rsyslogd)
  CGroup: name=systemd:/system/rsyslog.service
  └ 908 /sbin/rsyslogd -n -c 5

   However it says the log data should be sent to
   /tmp/var/log/messages  which does not exist. Of course there is
   /var/log/messages but I see nothing there. I guess the question
   then becomes should I create that file?

   Bob



--
http://www.qrz.com/db/W2BOD

box7

--
users mailing list
users@lists.fedoraproject.org
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


Re: syslogd in F-16 -

2012-04-21 Thread Reindl Harald


Am 21.04.2012 19:36, schrieb Bob Goodwin - Zuni:
> On 04/21/2012 01:19 PM, Todd Zullinger wrote:
>> Bob Goodwin - Zuni, Virginia, USA wrote:
>>> DD-wrt has a logging function which I am unable to make work per:
>>> http://www.dd-wrt.com/wiki/index.php/Logging_with_DD-WRT since it 
>>> apparently needs syslogd which has been
>>> replaced with syslog-ng if I understand what I've found.
>>
>> That page is referring to syslogd on the dd-wrt system, as far as I can 
>> tell.  It mentions in passing that if you
>> want to send logs to a remote host, that the remote host needs to be running 
>> a syslog daemon, but makes no
>> mention of what daemon that should be.  Any syslog daemon on the remote host 
>> would suffice.  I have rsyslog at
>> home with dd-wrt logging there.
>>
>Ok, I guess I didn't understand everything I read. It looks like
>I have rsyslog running.
> 
>However it says the log data should be sent to
>/tmp/var/log/messages  which does not exist. Of course there is
>/var/log/messages but I see nothing there. I guess the question
>then becomes should I create that file?

first please get rid of your strange message-style!

have you searched for rsyslog documentation?

per default rsyslog does not accept network connections
per default iptables does not open syslog port

both is a good and sane default, each service which
is not explicitly opened MUST NOT listen on the network

so you have some work todo on your fedora box
and "spit log to host xyz" is not enough!



signature.asc
Description: OpenPGP digital signature
-- 
users mailing list
users@lists.fedoraproject.org
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


help installing f16

2012-04-21 Thread Mike Wright

Hi all,

I downloaded the f16 install DVD (6.5 hours) and have been trying to 
install it to an lvm partition that is part of a Xen box.  Two days 
later I'm asking for help.


Hope this doesn't sound like a rant but I'm very frustrated.  This is a 
description of what I've been going through with a request for help at 
the very basic level of "Set Up your Drives".


I can't use livecd to install because that method won't install into a 
virtual environment and shutting down every other vm just so I can boot 
"bare metal" takes down my entire operation: no dns, no mail, no web, no 
firewall, no databases.  Nulla.  Nada.  Squatalucci.  Out of business 
while a livecd install takes place.  I'm not the only user here.  That 
means I take down every other business, too.


Wow!  Talk about regression.  I feel like I'm in the way-back machine or 
in need of invoking the "Help me, Mr. Wizard" invocation what with the 
ansi language and keyboard selectors.  Haven't seen those in many, many 
years, maybe a decade.  Then there is the disk setup.  Makes that good 
old p.o.s. called disk-druid look useful.


Totally screwed up the disk.  Replaced a functional msdos partitioning 
scheme with GUID partitioning that creates a header at the very end of 
the disk which renders fdisk useless and requires kpartx or parted to be 
used instead.


Problem is that kpartx is also a p.o.s. and parted is half brain dead.

e.g. parted requires start and end parameters for a partition but does 
not provide any info about the disk's layout.  Had to use the now 
deprecated fdisk to find that info.  Had to use dd to fill 10G with 
dev/zero to destroy the GPT but the new disk manager still managed to 
find a non-existent partition table which meant there was no free space 
on the empty drive.


OK, finally made it to the disk setup where the 2 drives, 10g and 1g, 
are detected and automatically placed into the right side box and the 
arrows for moving them left are greyed out and unusable.  Fine.  10g is 
the drive selected for the bootloader.  Press .




Created 10g mount point / ext4 and 1g swap.  Press . 
"Partitioning Errors.  You have not created a bootloader stage 1 target 
device".  Just peachy.  What does it want?  Where is that done?  Tried 
adding a /boot partition that will never be used but that, too, proved 
pointless.


Has anybody succeeded in installing from "install media" as opposed to 
"livecd media"?


Pointers?  Hand holding?  Advice?  etc.  (just don't rick roll me).

Thanks,
Mike Wright

p.s.  Brushing up on my cuneiform in anticipation of f22.  Practicing 
starting fires by smashing rocks together in anticipation of f30.  LOL

--
users mailing list
users@lists.fedoraproject.org
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


Re: help installing f16

2012-04-21 Thread Reindl Harald


Am 21.04.2012 21:02, schrieb Mike Wright:
> Hi all,
> 
> I downloaded the f16 install DVD (6.5 hours) and have been trying to install 
> it to an lvm partition that is part of
> a Xen box.  Two days later I'm asking for help.
> 
> Hope this doesn't sound like a rant but I'm very frustrated.  This is a 
> description of what I've been going through
> with a request for help at the very basic level of "Set Up your Drives".
> 
> I can't use livecd to install because that method won't install into a 
> virtual environment and shutting down every
> other vm just so I can boot "bare metal" takes down my entire operation: no 
> dns, no mail, no web, no firewall, no
> databases.  Nulla.  Nada.  Squatalucci.  Out of business while a livecd 
> install takes place.  I'm not the only user
> here. That means I take down every other business, too.

i stopped read here

if ANYTHING in a virtual machine brings the host down then
you have a much bigger problem as this guest - this has to
be impossible in a working virtualization!





signature.asc
Description: OpenPGP digital signature
-- 
users mailing list
users@lists.fedoraproject.org
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


Re: help installing f16 -> bott with "noefi" / "nogpt"

2012-04-21 Thread Reindl Harald


Am 21.04.2012 21:02, schrieb Mike Wright:
> Created 10g mount point / ext4 and 1g swap.  Press . "Partitioning 
> Errors.  You have not created a bootloader
> stage 1 target device".  Just peachy.  What does it want?  Where is that 
> done?  Tried adding a /boot partition that
> will never be used but that, too, proved pointless.
> 
> Has anybody succeeded in installing from "install media" as opposed to 
> "livecd media"?

anaconda is broken since many years in manual partitioning
and getting worser with each fedora release. AFAIK there is
a complete rewrite in progress

until that add "noefi" or "nogpt" as boot param
the need for a kernel-param to decide between DOS/GPT
is simply broken by design, but however you need it
_

the main problem here is the useless error message

a F15 install media has a different message where you can
successful find with google the "noefi" boot param for the
install DVD



signature.asc
Description: OpenPGP digital signature
-- 
users mailing list
users@lists.fedoraproject.org
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


Re: help installing f16

2012-04-21 Thread Mike Wright

On 04/21/2012 12:06 PM, Reindl Harald wrote:



Am 21.04.2012 21:02, schrieb Mike Wright:

Hi all,

I downloaded the f16 install DVD (6.5 hours) and have been trying to install it 
to an lvm partition that is part of
a Xen box.  Two days later I'm asking for help.

Hope this doesn't sound like a rant but I'm very frustrated.  This is a 
description of what I've been going through
with a request for help at the very basic level of "Set Up your Drives".

I can't use livecd to install because that method won't install into a virtual 
environment and shutting down every
other vm just so I can boot "bare metal" takes down my entire operation: no 
dns, no mail, no web, no firewall, no
databases.  Nulla.  Nada.  Squatalucci.  Out of business while a livecd install 
takes place.  I'm not the only user
here. That means I take down every other business, too.


i stopped read here

if ANYTHING in a virtual machine brings the host down then
you have a much bigger problem as this guest - this has to
be impossible in a working virtualization!


Thanks for your response, Reindl.

It is not a vm taking down the host.  It is I who must shutdown every vm 
in order to boot onto "bare metal" without a hypervisor.  That is 
required because livecd technology will not install into a virtual 
machine.  The only way I know to install into a vm is to use actual 
install media.

--
users mailing list
users@lists.fedoraproject.org
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


Re: help installing f16

2012-04-21 Thread Reindl Harald


Am 21.04.2012 21:15, schrieb Mike Wright:
> On 04/21/2012 12:06 PM, Reindl Harald wrote:
>>
>>
>> Am 21.04.2012 21:02, schrieb Mike Wright:
>>> Hi all,
>>>
>>> I downloaded the f16 install DVD (6.5 hours) and have been trying to 
>>> install it to an lvm partition that is part of
>>> a Xen box.  Two days later I'm asking for help.
>>>
>>> Hope this doesn't sound like a rant but I'm very frustrated.  This is a 
>>> description of what I've been going through
>>> with a request for help at the very basic level of "Set Up your Drives".
>>>
>>> I can't use livecd to install because that method won't install into a 
>>> virtual environment and shutting down every
>>> other vm just so I can boot "bare metal" takes down my entire operation: no 
>>> dns, no mail, no web, no firewall, no
>>> databases.  Nulla.  Nada.  Squatalucci.  Out of business while a livecd 
>>> install takes place.  I'm not the only user
>>> here. That means I take down every other business, too.
>>
>> i stopped read here
>>
>> if ANYTHING in a virtual machine brings the host down then
>> you have a much bigger problem as this guest - this has to
>> be impossible in a working virtualization!
> 
> Thanks for your response, Reindl.
> 
> It is not a vm taking down the host.  It is I who must shutdown every vm in 
> order to boot onto "bare metal" without
> a hypervisor.  That is required because livecd technology will not install 
> into a virtual machine.  The only way I
> know to install into a vm is to use actual install media.

i have not tried but how should the "live cd technology" even know
that it runs in a VM? however, live-cd-to-disk in my opinion is
generally unuseable due the lack of partitioning

as said in my following post: boot with "noefi" or/and "nogpt"
i needed this even with F15 on bare metal to get a working setup
with /boot on linux RAID1 and two RAID10 for system / data
which is definitly not possible in the default anaconda >= F15
while on F14 it was a normal thing like eat and drink

a really godd example that "improvements" in Fedora often are
horrible steps backwards and only improbements for
"click, click, OK" instalaltions



signature.asc
Description: OpenPGP digital signature
-- 
users mailing list
users@lists.fedoraproject.org
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


Re: help installing f16

2012-04-21 Thread Mike Wright

On 04/21/2012 12:39 PM, Reindl Harald wrote:



Am 21.04.2012 21:15, schrieb Mike Wright:

On 04/21/2012 12:06 PM, Reindl Harald wrote:



Am 21.04.2012 21:02, schrieb Mike Wright:



i have not tried but how should the "live cd technology" even know
that it runs in a VM?


I can't speak with authority but I think the livecd creates a mount 
point that occupies the same point required by an installcd, hence 
conflict.  Since it was there first the livecd wins.



however, live-cd-to-disk in my opinion is
generally unuseable due the lack of partitioning


maybe someday


as said in my following post: boot with "noefi" or/and "nogpt"
i needed this even with F15 on bare metal to get a working setup
with /boot on linux RAID1 and two RAID10 for system / data
which is definitly not possible in the default anaconda>= F15
while on F14 it was a normal thing like eat and drink

a really godd example that "improvements" in Fedora often are
horrible steps backwards and only improbements for
"click, click, OK" instalaltions


ROFL, my finger hurts already ;D
--
users mailing list
users@lists.fedoraproject.org
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


Re: help installing f16

2012-04-21 Thread Michael Hennebry

On Sat, 21 Apr 2012, Mike Wright wrote:

It is not a vm taking down the host.  It is I who must shutdown every vm 
in order to boot onto "bare metal" without a hypervisor.  That is 
required because livecd technology will not install into a virtual 
machine.  The only way I know to install into a vm is to use actual 
install media.


An iso is the standard method for VirtualBox.

--
Michael   henne...@web.cs.ndsu.nodak.edu
"On Monday, I'm gonna have to tell my kindergarten class,
whom I teach not to run with scissors,
that my fiance ran me through with a broadsword."  --  Lily
--
users mailing list
users@lists.fedoraproject.org
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