Mailinglist Archive: opensuse-project (142 mails)
| < Previous | Next > |
Re: [opensuse-project] End of life - what does it mean ?
- From: Christian Boltz <opensuse@xxxxxxxxx>
- Date: Fri, 18 Apr 2008 23:25:12 +0200
- Message-id: <200804182325.15697@xxxxxxxxxxxxxxx>
Hello,
on Freitag, 18. April 2008, Adrian Schröter wrote:
I think the best way is something between "do everything automatically
and perfect" and "ask the user to fix it" ;-)
Of course automatically updating all config files is nice, but I also
understand that it is not always possible or too difficult.
I'm pragmatic enough to prefer a 95% working update (with 5% "ask the
user to fix it") over a wannaby-perfect update that will never be
available because it takes too much time for the developers *g*.
I always install new openSUSE versions as updates and can happily report
that things usually just work[tm]. The only exception is my radeon
xinerama setup which is well-known by several X developers at
SUSE ;-) - however, this is caused by changes in the radeon driver and
not really SUSE's fault.
OK, I have to check some *.rpmnew or *.rpmorig files after the update
(hint: diff and patch are useful utilities!) but basically the update
never broke my system and it would also have worked without checking
the *.rpm* files.
*.rpm* files usually appear if a config file was manually changed, which
means the user/admin hopefully knows what he is doing. So he should
also know how to merge the configuration to the new config files.
For normal desktop users, this also means that *.rpm* files shouldn't be
a real problem because most desktop users don't change config files
in /etc.
...
Network installation / update work, at least for me ;-)
I updated a remote web server from 10.1 to 10.2 some months ago
(installation started with a bootloader entry, network installation
source and ssh -X to start YaST).
(It is still a good idea to have a recovery system available in case
something goes wrong. I didn't need it, but better safe than sorry...)
I even had a copy of the system running in a chroot, so the websites
were available during the update ;-) (with a minor downtime for
rebooting, mounting everything and starting the services in the
chroot). [1]
That said: The easier way to go would be updating the running system.
Basically this means that "zypper dup" and its graphical variant (which
is now called "Factory update") should be officially supported for
distribution updates.
Is this possible? (Un?)fortunately I have no idea how difficult this is
on the technical side. Feel free to enlighten me ;-)
Regards,
Christian Boltz
[1] Some notes for people who want to do this:
- mount --bind /proc, /sys and /dev into the chroot, otherwise many
things won't work
- choose the services to start carefully - for example, (re)starting
rcnetwork is not the best idea ;-) because the installation system
has already done that. Also avoid ssh and the firewall.
- starting daemons like Apache, MySQL etc. shouldn't be a problem
--
Früher mußte man den Müll heimlich im Wald verbuddeln;
Heute gibt es EBAY :-) [Axel Lindlau in suse-linux]
---------------------------------------------------------------------
To unsubscribe, e-mail: opensuse-project+unsubscribe@xxxxxxxxxxxx
For additional commands, e-mail: opensuse-project+help@xxxxxxxxxxxx
on Freitag, 18. April 2008, Adrian Schröter wrote:
On Wednesday 16 April 2008 16:15:06 wrote jdd:...
That is work more on *upgrading cleanly* a distro.
Sure, this is always wanted. But you need to take into account that
this is something we have only partly in our hands.
Enough open source projects do not care about this, so we need either
to write update scripts (to adapt a configuration for example) or ask
the users to fix it manually.
I think the best way is something between "do everything automatically
and perfect" and "ask the user to fix it" ;-)
Of course automatically updating all config files is nice, but I also
understand that it is not always possible or too difficult.
I'm pragmatic enough to prefer a 95% working update (with 5% "ask the
user to fix it") over a wannaby-perfect update that will never be
available because it takes too much time for the developers *g*.
I always install new openSUSE versions as updates and can happily report
that things usually just work[tm]. The only exception is my radeon
xinerama setup which is well-known by several X developers at
SUSE ;-) - however, this is caused by changes in the radeon driver and
not really SUSE's fault.
OK, I have to check some *.rpmnew or *.rpmorig files after the update
(hint: diff and patch are useful utilities!) but basically the update
never broke my system and it would also have worked without checking
the *.rpm* files.
*.rpm* files usually appear if a config file was manually changed, which
means the user/admin hopefully knows what he is doing. So he should
also know how to merge the configuration to the new config files.
For normal desktop users, this also means that *.rpm* files shouldn't be
a real problem because most desktop users don't change config files
in /etc.
...
I have a hosted server with 10.2, and little way to use anything
else (openSUSE). I would be much easier for me to use only one
distro, but I fear the online update...
k, but that will not help us to improve it ;)
Network installation / update work, at least for me ;-)
I updated a remote web server from 10.1 to 10.2 some months ago
(installation started with a bootloader entry, network installation
source and ssh -X to start YaST).
(It is still a good idea to have a recovery system available in case
something goes wrong. I didn't need it, but better safe than sorry...)
I even had a copy of the system running in a chroot, so the websites
were available during the update ;-) (with a minor downtime for
rebooting, mounting everything and starting the services in the
chroot). [1]
That said: The easier way to go would be updating the running system.
Basically this means that "zypper dup" and its graphical variant (which
is now called "Factory update") should be officially supported for
distribution updates.
Is this possible? (Un?)fortunately I have no idea how difficult this is
on the technical side. Feel free to enlighten me ;-)
Regards,
Christian Boltz
[1] Some notes for people who want to do this:
- mount --bind /proc, /sys and /dev into the chroot, otherwise many
things won't work
- choose the services to start carefully - for example, (re)starting
rcnetwork is not the best idea ;-) because the installation system
has already done that. Also avoid ssh and the firewall.
- starting daemons like Apache, MySQL etc. shouldn't be a problem
--
Früher mußte man den Müll heimlich im Wald verbuddeln;
Heute gibt es EBAY :-) [Axel Lindlau in suse-linux]
---------------------------------------------------------------------
To unsubscribe, e-mail: opensuse-project+unsubscribe@xxxxxxxxxxxx
For additional commands, e-mail: opensuse-project+help@xxxxxxxxxxxx
| < Previous | Next > |