How stable and usable is UFS2?

vertexSymphony

vertexSymphony

Hi people, so much time without reading the forums!
I have a question that I couldn't get answered by documentation :(

How stable is UFS with softupdates and journaling?
Is there any difference between 32 && 64 bits like zfs?

Thanks for your time.

Cheers, Alex.
 
It is the default FS for FreeBSD 9 (currently at RC-2), and it is considered stable for production. I don't have any experience in a server environment but I use it on my desktop with no issues at all. As a matter of fact it has saved me from many fscks.
I don't think that 32bit versus 64bit platform plays any role here.
 
xibo said:
UFS2 with soft updates is rock stable. I can't tell you about journaling though.
Click to expand...

Are you sure? Last time I tried I ended up on a horrible fsck thing when I rebooted and had to deal with Inodes and low level stuff for more than an hour.

It looked like a scene from a horror movie. Because I had a power outage during a heavy I/O operation.

P.S: I'm talking about +SJ in 9.x. Sorry for not clearing that up.
 
I think xibo means that it has not be tested enough in production environment compared to plain UFS2.
Journaling has been around for a while in CURRENT but only after 9-RELEASE we will be able to judge how it really behaves under load on certain type of controllers, in real environment.

George
 
SUJ makes fsck a lot better. I found out that the VirtualBox SATA emulation is not the greatest, and was amazed at how fast fsck completed. It's definitely one of the bigger features of 9.0.

As to how solid it is, can't say I've used it a lot, but it never had a problem in some very unplanned virtual hardware failures.
 
Thank you so much for the answers and for clearing up @xibo's comment.
I'll try doing some more testing with different setups.

Cheers.
 
I installed 9.0-BETA1 about 2 months ago on a HD that suffered a power fail on a 8.0-RELASE and data was lost. For the first month (september) I randomly unplugged power 2-3 times per day (brutally unplugged, pulling the cable) and never failed, never manually ran fsck, never start it in single user mode (and never read dmesg). No lost files, probably I lost some data in that files in some power fails, i.e. reverted to last good state, but never had HD failures as before (unreadable sectors, DMA unrecoverable errors).

In october I promoted it to 'production', i.e. I moved all data on the previous HD (8.2-RELEASE) but it still 9.0-BETA1, I'm waiting 9.0-RELEASE to freebsd-update it.

Code: 
uname -a
FreeBSD xxxxx.xxxxxxxxxx.xxx 9.0-BETA1 FreeBSD 9.0-BETA1 #0: Fri Sep  2 17:54:11 CEST 2011 
[email]root@xxxxx.xxxxxxxxxx.xxx[/email]:/usr/obj/usr/src/sys/XXXXX  i386

Code: 
uptime
3:23AM  up 21 days,  7:01, 1 user, load averages: 0.01, 0.06, 0.12

For me this is enough, without journaling the HD failure was about 1 on 10 on the same machine. The strange thing is that when the HD was lost, it had bad sectors, after a reinstallation the bad sectors disappeared. It happens to me 2 times on 8.x for sure plus another one on 7.x but I was too newbie (much more than now) to know what really happened.
 
freethread said:
I installed 9.0-BETA1 about 2 months ago on a HD that suffered a power fail on a 8.0-RELASE and data was lost. For the first month (september) I randomly unplugged power 2-3 times per day (brutally unplugged, pulling the cable) and never failed, never manually ran fsck, never start it in single user mode (and never read dmesg).
Click to expand...

That really puts a smile on my face, I have to say. Hours of rebuilding due to a power failure is not fun : Uninterruptible power supply not included
 
It is almost 2+ years I'm running FreeBSD on a couple of servers that had a few power failure (2 per year - due to electricians doing unauthorized works x( - UPS does not help in such case) and I've never experienced a disk failure, nor a single user mode reboot. I have to admit that I run also a gmirror on the disks, that makes me sleeping well.
I think UFS2 is really stable, maybe even more stable than how ZFS is, being the former a long-time technology implementation and the latter a more modern one.
 
freethread said:
No lost files, probably I lost some data in that files in some power fails, i.e. reverted to last good state, but never had HD failures as before (unreadable sectors, DMA unrecoverable errors).
Click to expand...
Those are low-level hardware problems and are not related to the filesystem resilience. They may occur due to the stress caused by the power failure itself.
 
A few occasions on ufs2 I have had unreadable files/dirs etc.

Which recovered after a reboot and/or fsck.

I dont know if they happened due to bad hdd, vnode issue etc.
 
chrcol said:
A few occasions on ufs2 I have had unreadable files/dirs etc.

Which recovered after a reboot and/or fsck.

I dont know if they happened due to bad hdd, vnode issue etc.
Click to expand...

Of course UFS2 is not black-magic! It cannot protect you against hardware problems.
 
Well, when I recently had the problem with the x11/nvidia-driver I suffered several lock-ups only recoverables by hard-reset buttoning. I can say they where not less than four in a row. However the UFS2+SUJ filesystem recovered like a charm every time.

I think one can trust it in production environments. Of course with appropriate backup policy, as always.
 
Thank you *so much* for sharing your experiences and knowledge !
I actually did some testing with emulators/virtualbox-ose (basically because how it handles sync) and some really heavy tests.
I can't say UFS2+SUJ is bulletproof (I only got a corrupted /etc/master.passwd. anyways, only took a moment to fix it), but indeed is much more stable than softupdates on his own.

Cheers.
 
fluca1978 said:
Of course UFS2 is not black-magic! It cannot protect you against hardware problems.
Click to expand...

correct, but because I said might it doesnt mean that was the fact. What I meant was it could have been due to bad hdd but the bad hdd was never confirmed afterwards, ie. it didnt blow up or anything and worked normally afterwards.

The most recent occurance tho I am fairly sure is due to a bad hdd. Had one last week, and is evidence pointing to one of the drives been bad.
 
You must log in or register to reply here.
Back
Top