Re: [Mageia-dev] ANN: Dracut now mounts /usr properly \o/

2011-12-16 Thread Thierry Vignaud
On 16 December 2011 01:41, Colin Guthrie mag...@colin.guthr.ie wrote:
 Don't we want to keep mkinitrd available as a fallback for some time?

 Maybe but for the next while we should force dracut on everyone and test
 that it works well. It's the only way we're going to be able to support
 systemd when LVM is used due to systemd needing info from udev and thus
 needing udev to be started in the initrd (even tho' udev is ultimately
 shut down and restarted, the metadata from the LVM layers started before
 persists due to /run/udev being shared and thus systemd deals with disk
 mounting fine.

Feel free to alter meta-task's prefer.vendor.list


Re: [Mageia-dev] ANN: Dracut now mounts /usr properly \o/

2011-12-16 Thread Anssi Hannula
On 16.12.2011 00:14, Colin Guthrie wrote:
 OK, so this took me a while to actually get around to setting up some
 VMs to test this. Sorry about that :s
 
 Anyway, I've tested with a / on ext4 + /usr on LVM + ext4.
 
 With appropriate fixes, I was able to generate an initrd without any
 specific config to tell dracut what to do. There is a caveat. /usr is
 mounted read only. I can tweak this, but by the same token I'm not sure
 I should... It's trivial to do a mount -o remount,rw /usr if you want
 to install packages etc. Otherwise, /usr should pretty much be read only
 anyway.. if something tries to modify stuff on /usr that's not part of a
 package installation process that is (IMO) a bug and should be fixed.

IMO init scripts should mount /usr as rw unless /etc/fstab instructs to
mount it as read-only.

 The code for mounting /usr is not the same as how it'll (eventually) be
 done upstream. It only deals with /usr and if you have e.g. /usr/bin on
 a separate partition they you are SOL and will have to wait for the
 upstream version.
 
 I think however that for the majority of setups the current solutions is
 likely good enough.
 
 
 Now, when I installed, dracut wasn't used as initrd, so if no one shouts
 too loudly I'll make it deprecate mkinitrd tomorrow and we can test
 installs where dracut generates the first initrd and see how it fairs.
 
 Please test and let me know. It's version 014.1mga2
 
 Col
 
 


-- 
Anssi Hannula


Re: [Mageia-dev] ANN: Dracut now mounts /usr properly \o/

2011-12-16 Thread Johnny A. Solbu
On Friday 16 December 2011 10:58, Anssi Hannula wrote:
 IMO init scripts should mount /usr as rw unless /etc/fstab instructs to
 mount it as read-only.

Agreed.
As far as I'm conserned, it should always comply with the instructions in 
/etc/fstab.

-- 
Johnny A. Solbu
PGP key ID: 0xFA687324


signature.asc
Description: This is a digitally signed message part.


Re: [Mageia-dev] ANN: Dracut now mounts /usr properly \o/

2011-12-16 Thread Colin Guthrie
'Twas brillig, and Thierry Vignaud at 16/12/11 09:29 did gyre and gimble:
 On 16 December 2011 01:41, Colin Guthrie mag...@colin.guthr.ie wrote:
 Don't we want to keep mkinitrd available as a fallback for some time?

 Maybe but for the next while we should force dracut on everyone and test
 that it works well. It's the only way we're going to be able to support
 systemd when LVM is used due to systemd needing info from udev and thus
 needing udev to be started in the initrd (even tho' udev is ultimately
 shut down and restarted, the metadata from the LVM layers started before
 persists due to /run/udev being shared and thus systemd deals with disk
 mounting fine.
 
 Feel free to alter meta-task's prefer.vendor.list

That will only affect new installs right? I want to make sure that I
force it on everyone hence why I was thinking of obsoleting, but perhaps
I should just put it in as a require somewhere (e.g. to systemd) such
that everyone gets it but can still technically use mkinitrd should they
have a real need (i.e. dracut is really broken!)

Col

-- 

Colin Guthrie
colin(at)mageia.org
http://colin.guthr.ie/

Day Job:
  Tribalogic Limited http://www.tribalogic.net/
Open Source:
  Mageia Contributor http://www.mageia.org/
  PulseAudio Hacker http://www.pulseaudio.org/
  Trac Hacker http://trac.edgewall.org/


Re: [Mageia-dev] ANN: Dracut now mounts /usr properly \o/

2011-12-16 Thread Thierry Vignaud
On 16 December 2011 13:54, Colin Guthrie mag...@colin.guthr.ie wrote:
 Maybe but for the next while we should force dracut on everyone and test
 that it works well. It's the only way we're going to be able to support
 systemd when LVM is used due to systemd needing info from udev and thus
 needing udev to be started in the initrd (even tho' udev is ultimately
 shut down and restarted, the metadata from the LVM layers started before
 persists due to /run/udev being shared and thus systemd deals with disk
 mounting fine.

 Feel free to alter meta-task's prefer.vendor.list

 That will only affect new installs right?

Of course

 I want to make sure that I
 force it on everyone hence why I was thinking of obsoleting, but perhaps
 I should just put it in as a require somewhere (e.g. to systemd) such
 that everyone gets it but can still technically use mkinitrd should they
 have a real need (i.e. dracut is really broken!)


Re: [Mageia-dev] ANN: Dracut now mounts /usr properly \o/

2011-12-15 Thread Colin Guthrie
'Twas brillig, and Olivier Blin at 15/12/11 23:36 did gyre and gimble:
 Colin Guthrie mag...@colin.guthr.ie writes:
 
 Now, when I installed, dracut wasn't used as initrd, so if no one shouts
 too loudly I'll make it deprecate mkinitrd tomorrow and we can test
 installs where dracut generates the first initrd and see how it fairs.
 
 Don't we want to keep mkinitrd available as a fallback for some time?

Maybe but for the next while we should force dracut on everyone and test
that it works well. It's the only way we're going to be able to support
systemd when LVM is used due to systemd needing info from udev and thus
needing udev to be started in the initrd (even tho' udev is ultimately
shut down and restarted, the metadata from the LVM layers started before
persists due to /run/udev being shared and thus systemd deals with disk
mounting fine.

So while we can theoretically keep mkinitrd, I've very much like to see
if we can get away with just having to support one initramfs generator
if we can possibly get away with it.

Col


-- 

Colin Guthrie
colin(at)mageia.org
http://colin.guthr.ie/

Day Job:
  Tribalogic Limited http://www.tribalogic.net/
Open Source:
  Mageia Contributor http://www.mageia.org/
  PulseAudio Hacker http://www.pulseaudio.org/
  Trac Hacker http://trac.edgewall.org/