How to EXT4

[FeatherMonkey]

Now I've only tested this in a virtual environment I take no blame if you decide to try something experimental and it breaks. So I wouldn't try it if you care about the data and don't have a backup, Ideally a clone of ext3 as AFAIK it can't be undone, though who know perhaps someone will correct me.

How to Ext4

For 11.1

1st we change the options and flags

tune2fs -O extents,uninit_bg /dev/ROOT
tune2fs -O extents,uninit_bg /dev/HOME

2nd open /etc/fstab with the editor of your choice and change ext3 to ext4dev

Quick way
sed -i 's/ext3/ext4dev/g' /etc/fstab
check
cat /etc/fstab

3rd
Add ex4dev to INITRD_MODULES in /etc/sysconfig/kernel

4th now to make the the initrd support it..
sudo mkinitrd

5th We now need to run e2fsck
You can either go to single user mode
init 1
Or a reboot I tried a reboot seemed fine grumbled a little but did it fine.

How to Ext4

For 11.2

1st we change the options and flags

tune2fs -O extents,uninit_bg /dev/ROOT
tune2fs -O extents,uninit_bg /dev/HOME

2nd open /etc/fstab with the editor of your choice and change ext3 to ext4

Quick way
sed -i 's/ext3/ext4/g' /etc/fstab
check
cat /etc/fstab

3rd
Add ex4 to INITRD_MODULES in /etc/sysconfig/kernel

4th now to make the the initrd support it..
sudo mkinitrd

5th We now need to run e2fsck
You can either go to single user mode
init 1
Or a reboot I tried a reboot seemed fine grumbled a little but did it fine.

If you do get it failing on a system either / or /home as long as mkinitrd run properly then you should have fsck.ext4/dev I had the 11.2 not do /home and 11.1 not do root.

I found one other snippet as to whether it works <shrug> but then I wouldn't be doing this to a system that I cared about.
Append this to the /boot/grub/menu.lst on the kernel line
rootflags=data=ordered
related to Ext4 data loss; explanations and workarounds - News - The H Open Source: News and Features fix found(Maybe, don't tell me if you get data loss) Ext4 - ArchWiki

All info extracted from Ext4 Howto - Ext4 after hint from microchip 11.2 Milestone1 Initial Comments & Issues - openSUSE Forums

[BenderBendingRodriguez]

I do not recommend running 11.1 with the ext4dev. It was marked as experimental for a purpose. The chances to lose data are not worth the performance benefits.

[FeatherMonkey]

I do not recommend running ext4 on either if you care about data loss, certainly not without a backup.

[BenderBendingRodriguez]

I certainly disagree with You, i've been using ext4 with 11.2 M2 and didn't encounter any data loss. The problem of losing data is caused not by a BUG but by the programs, that has been talked about for months already. I guess the only data you can lose is the one that is actually opened and when you don't install any applications then you may only lose some settings in KDE.

[microchip8]

Originally Posted by FeatherMonkey

I do not recommend running ext4 on either if you care about data loss, certainly not without a backup.

The data loss problems have been fixed by Ted and fixes will be present in 2.6.30. There is no reason why not to use Ext4

[BenderBendingRodriguez]

While i agree it is fixed i disagree that this is a problem because delayed allocation is a feature and what Ted wrote is a "workaround" for applications that uselessly write small files so many times.

[FeatherMonkey]

Well I can't comment but if you follow the link after the supposedly fix of appending to the kernel line, I don't see why it shouldn't work(Seems to be forcing some mode over a default). If my understanding is correct it was design and the fix is yet to be back ported, reading the article implies the fix has only just made the 2.6.30.

But I don't follow fs or kernel mailing lists.. and I certainly don't know what the Suse team are backporting. The changelog for the 11.1 source is still showing relatively recent ext4 patch fixing, so it seems it is still being debugged.

[microchip8]

Originally Posted by BenderBendingRodriguez

While i agree it is fixed i disagree that this is a problem because delayed allocation is a feature and what Ted wrote is a "workaround" for applications that uselessly write small files so many times.

Not just the workaround. There were bugs present in the delayed allocation code itself which have been fixed now. Yes, programs are often coded bad, especially when it comes to fsync() calls for flushing data to disk

[BenderBendingRodriguez]

Yes, that's what she said

I just don't want everyone get on the ext4 wagon and then complain that they lost data etc. , only experienced users should attempt to do this or the ones that know what they are doing instead of simply following some instructions

By the way, that upgrade enables all those "new" features (specifically extents) only for new files so i recommend to make an unconditional update that will replace all the system files thus enabling extents for them.

[1michael1]

Originally Posted by BenderBendingRodriguez

Yes, that's what she said

I just don't want everyone get on the ext4 wagon and then complain that they lost data etc. , only experienced users should attempt to do this or the ones that know what they are doing instead of simply following some instructions

By the way, that upgrade enables all those "new" features (specifically extents) only for new files so i recommend to make an unconditional update that will replace all the system files thus enabling extents for them.

BenderBendingRodriguez,
I agree with you