You are not logged in.
It's not hellbent belief, it's knowledge based on experience. On perfectly healthy hard drives, I've had JFS lose written files after forced reboots many, many times, when other filesystems did not. I'd have X freeze up and have to hit the interrupt switch, and when I'd reboot, half the man pages would be gone, or /var/lib/pacman, or all the text files in /etc.
Offline
Seems that JFS raises most questions, since it is my favorite let me share some success and personal opinions:
- my home computer has jfs on all partitions, including / and /boot (with lilo, old habits die hard).
- I'm in charge of a web server delivering an average 50GB/day from an external iscsi device (yes, it's Arch), jfs formated. There is also an mysql server with the databases on the iscsi storage, also jfs, several hundreds of queries per second. Soon this setup will celebrate 1 year without any filesystem problems.
- I also run smaller servers, no problems with jfs, but much less load.
- I used reiserfs many years with little problems, but I lost once some files after repeated power failures.
- ext3 on Fedora (FC3 and 4 I think...) was a nightmare. I let the system boot after a failure, it said it was going to fix the filesystem an I let it do so... /etc/fstab turned into a binary file after that. Many other files where completely damaged. This happened twice.
- I only set up one xfs system. Pacman is obviously much slower than on similar computers with jfs or reiserfs.
XFS seems targeted to big (realy big) machines with big ups backup, but has a lot of utilities. Reiserfs is in a dead end and does not have a big company behind. Ext3 has his limits because of its, well..., primitive roots.
Frankly I think any fs in kernel is good enough. After all the decision is yours, it's like choosing a distro, they are all a linux kernel with a GNU environment, more or less.
Offline
This is a great thread as I am waiting to receive my new laptop and will be deciding what file system to use. I have never really learned much about file systems. I have always used ext3 or reiserfs with no problems.
Offline
I use xfs, never had any problems, even after power outage recovery was as fast as default boot.
however i don't use default format settings.
These are mine:
mkfs.xfs -l version=1,internal,size=128m -i size=2048,attr=2 -d unwritten=0,agcount=4 -f /dev/sda2
to prevent fragmentation completely I have additional flags in fstab
xfs defaults,nodiratime,noatime,logbufs=8,biosize=16,allocsize=512m
so after two yrs
fragmentation factor 0.58%
I don't suggest xfs but I really like it
Offline
I live in an area where there are frequent power outages. I have set up my system using Ext3 with full journaling enabled and have not noticed any loss in performance in fact subjectively the system seems much more responsive. The up side is that even though I have not used a UPS I have not lost data over countless unexpected shutdowns.
Check out this thread http://bbs.archlinux.org/viewtopic.php?id=12865
Last edited by kishd (2007-12-29 18:38:10)
---for there is nothing either good or bad, but only thinking makes it so....
Hamlet, W Shakespeare
Offline
Great thread! I decided to go with JFS.
So my next question.
Lot's of people talk about deadline scheduler. As far as I understand advise to use deadline shcedule comes from times when default scheduler was some not good scheduler. Now kernel comes with new default scheduler - CFS.
Should'nt CFS be better than deadline ?
If I change scheduler does it affects only FS or all running applications ?
Can I use different scheduler for FS, and different one for everything else ?
What is the visible difference between deadline and CFS for end user ?
Offline
these are I/O schedulers, not sure what "everything else" do you mean?
you can set elevator cfq or as or deadline.
I would suggest either as or cfq Currently these seems best for mixed workload (desktop).
You can test each scheduler easily by appending elevator to the kernel line in your menu.lst
Offline
these are I/O schedulers, not sure what "everything else" do you mean?
I'm a noob, so forgive me if I do not know what the heck a scheduler is. I thought it is a thread/process switching scheduler. So it affects all applications. I do not know what an IO scheduler is. I probably need to read more on that
EDIT: I just read wikipedia on IO schedulers. Never mind my stupid question. I confused CFS which is a task scheduler with deadline scheduler which is IO scheduler.
Last edited by vagif (2007-12-29 20:36:27)
Offline
I got a new hard drive for christmas and found this thread very helpful.. I'm just wondering, what is considered a large file (thinking about using XFS)? I suppose video files at 150mb are qualified?
Last edited by burk (2007-12-30 11:15:21)
Offline
xfs is optimized for multimedia more than any other fs
Offline
One quick question, but do you use JFS for your / partition as well? or do you use a combination of different file systems?
I have a small /boot with ext2. All other partitions be jfs. I'm not a fan of mixing file systems instead of it could be a good idea to use different for at example /var and /home.
I think any fs in kernel is good enough. After all the decision is yours, it's like choosing a distro, they are all a linux kernel with a GNU environment, more or less.
That is the best summary. Where have i to sign?:)
One other thing before i forgot it: Thanks PDExperiment626 for the fantastic wiki about jfs.
Last edited by attila (2007-12-30 09:22:37)
Offline
@attila: Thanks for your input to my question! I'm actually trying out different setups on a spare box just to see what might be best/optimal configuration might be.
If I hit on anything interesting, I'll post my findings in these forums :-)
Offline
I use xfs, never had any problems, even after power outage recovery was as fast as default boot.
however i don't use default format settings.
These are mine:
mkfs.xfs -l version=1,internal,size=128m -i size=2048,attr=2 -d unwritten=0,agcount=4 -f /dev/sda2to prevent fragmentation completely I have additional flags in fstab
xfs defaults,nodiratime,noatime,logbufs=8,biosize=16,allocsize=512m
Uhh, broch, I really hope you meant "512k" not "512m"... Because it is possible to allocate 512 MB for growing files, and it isn't pretty.
Offline
my mistake
however for MythTV
http://s.webarmadillo.net/wordpress/?p=12
http://gentoo-wiki.com/HOWTO_Setup_Kern … for_MythTV
512m
Last edited by broch (2008-01-04 00:51:53)
Offline
It's also worth setting lazy-count to 1 when using mkfs.xfs, since recently.
Offline
Large files usually refer to the Gigabytes' range, especially so for XFS since the old Y2K FS benchmarks. Now it's probably in the terra's. I don't know how effective XFS can be with the average Joe's multimedia files since I remember having read it wouldn't make sense for a Desktop home computer. I'd guess portions on the disk where a lot of writes occur are the best places to put XFS on?
I don't think it's alright to take authority over something based on just experience alone. The comments on JFS have not been just, and are a little unfair. True, I have experienced the cons more than 1 time (I do have a sane system though), but I also remember having done something out of ordinary before the incidents. For now, everything is good and going strong and I hope it remains so for every other JFS user
I need real, proper pen and paper for this.
Offline
Just for the record, I'd like to chime in with my support for JFS. I've been running it as my FS of choice for quite some time now, and I've never encountered any problems.
The thing I don't like about ext3 is that you have to mess about with all of these options when you make a filesystem in order to make sure it doesn't spend an hour running fsck every 99 boots or whatever. With JFS, you just point it at a partition and straight away you've got a decent, well-performing journaled FS.
If you want to point the finger anywhere safety-wise, I believe that ReiserFS (v3) is particularly susceptible to data loss. I can't find the link ATM, but apparently on sudden power loss, the RAM is pretty much the first thing to suffer. If the ReiserFS driver reads from the corrupted RAM, it may in turn corrupt the FS superblock, and POW! all of your data is gone, since the superblock cannot be restored. I think the thing that makes this specific to ReiserFS is that with other FSes, it is possible to restore the superblock by churning through all of the data on the HDD and retrieving the lost files, but the way ReiserFS stores the files makes this impossible.
Having said all of the above, I actually used ReiserFS for years and never lost any data, even with lots of sudden power losses.
Also, I don't think anyone's yet mentioned that the JFS fsck utility is really nice and gives lots of useful output.
Offline
Offline
I don't think anyone's yet mentioned that the JFS fsck utility is really nice and gives lots of useful output.
And it's fast!
I need real, proper pen and paper for this.
Offline
Since i started this post, i just want to say, that I put JFS on my laptop.
So far works great.
Offline
I use ext3 and have been experience many problems, but I think it is because my hard drive is Y2K quality... yeah that might be it.
Last edited by theringmaster (2008-01-10 00:42:54)
Check me out on twitter!!! twitter.com/The_Ringmaster
Offline