Which File System do you predominately use?
Submitted by busfault on Fri 14 Dec 2007
| MINIX |
   3% | 53 votes |
| ext2 |
1% | 27 votes |
| ext3 |
70% | 1094 votes |
| Reiser |
13% | 208 votes |
| JFS |
1% | 20 votes |
| XFS |
8% | 136 votes |
| other |
0% | 15 votes |
| Total 1554 votes |
Posted by dkg (216.254.xx.xx) on Fri 14 Dec 2007 at 16:55
[ Send Message | View dkg's Scratchpad | View Weblogs ]
[ Send Message | View dkg's Scratchpad | View Weblogs ]
Has anyone started experimenting with ext4 yet? If you have, what do you think of it? Do you have the amount of disk space to make it worthwhile?
[ Parent ]
Like most people here I use ext3 on Linux/pc hardware. I mostly use ext3 on raw disk partitions, but going forward I expect to migrate mostly to LVM2 for all my systems.
On paper XFS is a pretty good filesystem, it's apparently very fast, with lots of features. I gather there are issues with it that put people off. Does anyone other than SGI use it?
At work our AIX boxes run on JFS2 on IBM LVM which is a bit easier than on PC hardware as the System P firmware understands LVM/JFS so you don't need to mess about with a non-LVM /boot.
--
"It's Not Magic, It's Work"
Adam
On paper XFS is a pretty good filesystem, it's apparently very fast, with lots of features. I gather there are issues with it that put people off. Does anyone other than SGI use it?
At work our AIX boxes run on JFS2 on IBM LVM which is a bit easier than on PC hardware as the System P firmware understands LVM/JFS so you don't need to mess about with a non-LVM /boot.
--
"It's Not Magic, It's Work"
Adam
[ Parent ]
I marked ext3 on the poll, but I've got a combination of LVM, ext3, and XFS on our new file server. I wouldn't have put XFS on at all, but ext3 has an 8TB filesystem limit.
(Yes, there's a bit of grandstanding in that paragraph, but it's offset by the fact that when anyone else goes to that large a volume, you'd probably want to know that ext3 just isn't going to work out of the box on stock Etch.)
[ Parent ]
I use ext3 for the / partition (around 200 Mg) because grub doesn't like XFS, for all other partitions i use XFS on worksations and servers.
I'm very happy with it, no problems so far, once in a wile i run xfs_fsr to keep it at max performance.
I'm very happy with it, no problems so far, once in a wile i run xfs_fsr to keep it at max performance.
[ Parent ]
I see:
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=243835
https://bugs.launchpad.net/debian/+source/grub/+bug/8058
I'm trying out a Debian install in a VBox container at the moment to see if this is fixed yet. The Etch installer picked LILO for me automatically I wasn't offered GRUB as an option and it rebooted perfectly after the install. Now I have to see it I can put GRUB on and get it to work...
--
"It's Not Magic, It's Work"
Adam
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=243835
https://bugs.launchpad.net/debian/+source/grub/+bug/8058
I'm trying out a Debian install in a VBox container at the moment to see if this is fixed yet. The Etch installer picked LILO for me automatically I wasn't offered GRUB as an option and it rebooted perfectly after the install. Now I have to see it I can put GRUB on and get it to work...
--
"It's Not Magic, It's Work"
Adam
[ Parent ]
Got GRUB working, but I had to hack the installer from SID:
http://lists.alioth.debian.org/pipermail/pkg-grub-devel/2005-May/ 000945.html
--
"It's Not Magic, It's Work"
Adam
http://lists.alioth.debian.org/pipermail/pkg-grub-devel/2005-May/ 000945.html
--
"It's Not Magic, It's Work"
Adam
[ Parent ]
I'm using XFS over LVM since two years and I never experienced a problem with it !
preferred feature : extend while mounted
I used reiserfs before which can also be reduced unmounted (XFS can't, you must copy the fs and recreate a smaller one).
I switched to XFS definitely when I saw reiserfs taking more than 30s mounting a 160+Gb fs !!!
quotas :
xfs(yes) reiserfs(no) ext3(yes)
no more lost+found in each mount point :
xfs(yes) reiserfs(yes) ext3(no)
no need to be checked :
xfs(yes) reiserfs(yes) ext3(no)
And more and more features I forgot...
preferred feature : extend while mounted
I used reiserfs before which can also be reduced unmounted (XFS can't, you must copy the fs and recreate a smaller one).
I switched to XFS definitely when I saw reiserfs taking more than 30s mounting a 160+Gb fs !!!
quotas :
xfs(yes) reiserfs(no) ext3(yes)
no more lost+found in each mount point :
xfs(yes) reiserfs(yes) ext3(no)
no need to be checked :
xfs(yes) reiserfs(yes) ext3(no)
And more and more features I forgot...
[ Parent ]
Posted by Anonymous (86.112.xx.xx) on Thu 20 Dec 2007 at 23:42
ext3 also allows you to extend whilst the partition is mounted - under LVM at least. Not sure about natively.
I use ext3 not because it is the best, but because it is the default. It is "good enough", and everybody understands it. If there are problems with my server(s) it is unlikely to be ext3 related - stable and well-known are two terms that are good for servers..
I use ext3 not because it is the best, but because it is the default. It is "good enough", and everybody understands it. If there are problems with my server(s) it is unlikely to be ext3 related - stable and well-known are two terms that are good for servers..
[ Parent ]
Used Reiser a while ago, but I feel safer with my data on ext3. There's plenty of support for ext3 with distributions, sometimes other filesystems feel like second best.
One of my concerns with Reiser at the time was a Reiser disk image on a Reiser filesystem could cause problems with a fsck if a tree rebuild was needed. As for XFS, it looks nice but aggressively caches data, so I'd prefer it on a server with a UPS.
Stick with the most safe filesystem possible unless you really need the extra performance.
One of my concerns with Reiser at the time was a Reiser disk image on a Reiser filesystem could cause problems with a fsck if a tree rebuild was needed. As for XFS, it looks nice but aggressively caches data, so I'd prefer it on a server with a UPS.
Stick with the most safe filesystem possible unless you really need the extra performance.
[ Parent ]
Previously XFS, now always JFS which is, from my experiences, currently the best performing file system around for Lnx. I also didn't encountered any problems with it in the last 2-3 years that I'm using it with Lnx. Check also this site for performance benchmarks:
http://linuxgazette.net/102/piszcz.html
At work (that is, our customers... :-) JFS2 (AIX) is the only answer, sometimes replaced with Veritas VxFS, mostly in combination with Veritas cluster. I haven't tried Sun's ZFS though, which is supposed to be a step forward in a FS devel, although I suspect that there are just a bit too many workarounds used. Notably that statement comes from Sun itself, but still it is worthy testing:
http://www.sun.com/2004-0914/feature/
http://digitalbadger.net/archives/35-ZFS-and-caching-for-performa nce.html
Speaking however, enterprise stuff, GPFS beats all others, although it differs from a traditional fs concept:
http://www.research.ibm.com/haifa/projects/storage/gpfs.html
http://www.betanews.com/article/IBM_Builds_Super_Fast_File_System /1141943597
http://arstechnica.com/news.ars/post/20060310-6362.html
http://linuxgazette.net/102/piszcz.html
At work (that is, our customers... :-) JFS2 (AIX) is the only answer, sometimes replaced with Veritas VxFS, mostly in combination with Veritas cluster. I haven't tried Sun's ZFS though, which is supposed to be a step forward in a FS devel, although I suspect that there are just a bit too many workarounds used. Notably that statement comes from Sun itself, but still it is worthy testing:
http://www.sun.com/2004-0914/feature/
http://digitalbadger.net/archives/35-ZFS-and-caching-for-performa nce.html
Speaking however, enterprise stuff, GPFS beats all others, although it differs from a traditional fs concept:
http://www.research.ibm.com/haifa/projects/storage/gpfs.html
http://www.betanews.com/article/IBM_Builds_Super_Fast_File_System /1141943597
http://arstechnica.com/news.ars/post/20060310-6362.html
[ Parent ]
Posted by Anonymous (68.45.xx.xx) on Sun 23 Dec 2007 at 07:21
I love XFS, I found parameters to really make it fly. For formatting I use mkfs.xfs -f -L my_filesystem_name -ssize=4096 -l size=64m,sunit=512 /dev/sda1
and when mounting i use these options: defaults,noatime,nodiratime,logbufs=8 0 0
It really makes the FS fly in everything, not just removing files :)
and when mounting i use these options: defaults,noatime,nodiratime,logbufs=8 0 0
It really makes the FS fly in everything, not just removing files :)
[ Parent ]