EXT4 on CentOS 5.2 ?

[mrtdc]

I've been trying to determine if there is a stable / straightforward way to run EXT4 filesystems with CentOS 5.2 in production, now that ext4 is apparently released into stable production in the linux kernel. I assume it will require a kernel module at the very least, and possibly a new kernel entirely - but I was wondering if this is a topic that has been addressed and documented elsewhere already, or if it is still bleeding edge / not recommended / not well integrated.

I ask in particular because I have a D2D backup server project, in pre-deployment right now, using XFS as the filesystem for a large raid6 disk array (slightly larger than 12 Tb usable capacity). The volume size is such that EXT3 is not an option. XFS seems to work fine, and is easily supported now in CentOS5.2 which is great. Alas I've learned the hard way with a recent hard power outage on this server, that XFS appears slightly less tolerant of unclean shutdowns than EXT3 style filesystems.

Since performance isn't a big issue on this - it is a backup server, not production users hammering data - having a more robust / forgiving filesystem might be a nice thing, if possible. Hence my thoughts on EXT4.

Any pointers or comments certainly are greatly appreciated,


Tim Chipman

[AlanBartlett]

From the 2.6.28 kernel sources --

[code]
CONFIG_EXT4_FS:
│
│ This is the next generation of the ext3 filesystem.
│
│ Unlike the change from ext2 filesystem to ext3 filesystem,
│ the on-disk format of ext4 is not forwards compatible with
│ ext3; it is based on extent maps and it supports 48-bit
│ physical block numbers. The ext4 filesystem also supports delayed
│ allocation, persistent preallocation, high resolution time stamps,
│ and a number of other features to improve performance and speed
│ up fsck time. For more information, please see the web pages at
│ http://ext4.wiki.kernel.org.
│
│ The ext4 filesystem will support mounting an ext3
│ filesystem; while there will be some performance gains from
│ the delayed allocation and inode table readahead, the best
│ performance gains will require enabling ext4 features in the
│ filesystem, or formating a new filesystem as an ext4
│ filesystem initially.
[/code]
With regards to [i]CentOS 5[/i], the [i]ext4[/i] file system [i]is[/i] bleeding edge, is not integrated and is not recommended. If you do attempt it, you will: (1) be on your own, (2) will need a custom kernel and many fs utilities rebuilt.

If you must have ext4, please wait for [i]CentOS 6[/i] and ask again . . .

My advice is to invest in a good quality UPS that will provide sufficient power to enable a clean shutdown to take place once a mains outage has been signalled and to stay with your current file system.

[PRandal]

ext4 is available as a "technology preview" in RedHat 5.3, released today:

"The latest generation of the ext filesystem, ext4, is available in this release as a Technology Preview. Ext4 is an incremental improvement on the ext3 file system developed by Red Hat and the Linux community. The release name of the file system for the Technology Preview is ext4dev.
The file system is provided by the ext4dev.ko kernel module, and a new e4fsprogs package, which contains updated versions of the familiar e2fsprogs administrative tools for use with ext4. To use, install e4fsprogs and then use commands like mkfs.ext4dev from the e4fsprogs program to create an ext4-base file system. When referring to the filesystem on a mount commandline or fstab file, use the filesystem name ext4dev."

So it'll be in CentOS 5.3.

Phil

[AlanBartlett]

Thanks for the update, [b]Phil[/b]. Can we put you down as being prepared to answer all [i]ext4[/i] fs queries once [i]update three[/i] is released? :-P

[scottro]

Ah, this goes back to the other thread, why one might want to have a separate boot partition. The grub that CentOS and Fedora uses can't boot ext4fs. (Though there might be patches, but at present, only Ubuntu's jaunty is shipping with a grub able to handle ext4fs.)

With a separate /boot partition, this becomes a non-issue. Without it, well--one would have to repartition, add an ext2 or 3 /boot, then redo various and sundry files.

[TonyG2]

The question I have on the EXT4 code in the just release 5.3 kernel is this- How "stable" is it?

The question I'm really asking is how close is the code in the CentOS 5.3 fs/ext4/ kernel tree to that in Linux 2.6.28.1 ? I've played with the latter and like what I see performance-wise. It seems to be about 50% better on my workload than a similarly configured EXT3 filesystem! If I create the EXT4 f/s with e2fsprogs 1.43-3 (+ some patches I developed to allow explicit journal file positioning), the 5.3 ext4dev.ko module will not have a bar of it. I can recall reading somewhere about a development 'test' flag that might be required but the error indicates that the code in the 5.3 release is considerably different to that in 2.6.28.1 and I'd be largely wasting my time continuing down this path...

Tony

[justinc]

I'm trying to mount my Ubuntu 9.04 on ext4 with e4fsprogs and it isn't working.

[root@echelon ~]# mount -t ext4dev /dev/sda1 /mnt/ubuntu/
mount: wrong fs type, bad option, bad superblock on /dev/sda1,
missing codepage or other error
In some cases useful info is found in syslog - try
dmesg | tail or so


/var/log/messages

Jun 9 09:37:06 echelon kernel: EXT4-fs: sda1: not marked OK to use with test code.


:-(

[cry4dawn]

what is the status of ext4 in Centos 5.4 at this point. I would love to test it as a main filesystem (/) as opposed to a partition on a secondary disk as I have been doing.
my atypical layout is
/boot 200M ext2
8000M swap
/ * ext3

would I be able to convert / to ext4 and be relativly safe at this point?

[fugtruck]

I'm using it on a number of production servers and have had no problems. I don't have it installed on any root or boot partitions, just /var and different data partitions.

[cry4dawn]

How did you convert /var?
Link would be fine as the ones I have found don't work with selinux enabled.