Ext4 data loss; explanations and workarounds

H_TeXMeX_H · 03-19-2009, 02:13 PM

http://www.h-online.com/open/Ext4-da...--/news/112892

To prevent data loss on ext4, the main idea is:

Quote:

Nevertheless, as a workaround, he very quickly wrote patches for Ext4 that recognise the rename() situation and ensure it behaves like Ext3, and a second procedure that uses ftruncate(). Now, however, he has provided a "proper" solution. The new ext4 mounting option "alloc_on_commit" gives Ext4 a mechanism analogous to "data=ordered" in Ext3, whereby metadata is not committed in the journal until after blocks have been allocated and the data has been written. However, this change probably won't arrive until version 2.6.30 of the kernel at the earliest.

yet another article one it, maybe a bit clearer:
http://thunk.org/tytso/blog/2009/03/...-file-problem/

mostlyharmless · 03-20-2009, 03:22 PM

Not that I know anything about it, but I find the comment that

Quote:

Ext4 behaves precisely as demanded by the POSIX standard for file operations

and that

Quote:

the problem lies with application developers who take the forgiving behaviour of Ext3 to be a standard.

rather disingenuous.
What are we to conclude from this? "It's not my fault, in fact it isn't a problem, and if it were it'd be POSIX's fault?"

This is the kind of thing we're used to hearing from Microsoft about a perceived shortcoming, like the UAC in Vista, really being a good thing.

Guess I'm just a naive user (of ext3, mostly) but integrity of the filesystem in the event of a system crash is probably the most important feature.

H_TeXMeX_H · 03-20-2009, 04:40 PM

Perhaps the designer / developer of ext4 should have thought a bit more into the problem.

ssenuta · 03-20-2009, 11:50 PM

I am planning to install Mandriva-2009.1 with the Ext4 fs when it becomes available. Mandriva will probably not have kernel-2.6.30 because it isn't even available yet. So, to protect against data loss, I was going to mount my Ext4 partitions with the -o nodelalloc option. I thought this was the option that enabled Ext4 to behave like Ext3 & write data to the disk without the Delayed Allocation feature.

I am now uncertain what option I should use.
1.) mount -t ext4 /dev/sdb1 / -o nodelalloc
2.) mount -t ext4 /dev/sdb1 / -o alloc_on_commit

Can anyone explain the difference?

I could be wrong about the "nodelallac" option because I found out about it be reading some pretty technical Ext4 internet postings.The discussions there were way above my head, so maybe I just didn't comprehend what was said properly.

syg00 · 03-21-2009, 12:06 AM

Quote:

Originally Posted by H_TeXMeX_H

Perhaps the designer / developer of ext4 should have thought a bit more into the problem.

I suspect he probably did - seeing as how he is also responsible for ext2 (as we know it) and ext3 (go to the bottom of "man mkfs.ext3" and see if the name looks familiar).
New features expose new problems - nothing remarkable about that. Unfortunately this came up so late in the release cycle. Personally I can't understand why people rush into the "latest and greatest" filesystem - it's what holds all your data.
Bragging rights about being on the "bleeding edge" lose their gloss when your data is trashed.

My $AU0.02

H_TeXMeX_H · 03-21-2009, 04:37 AM

Quote:

Yet another solution would be to mount ext4 volumes with the nodelalloc mount option. This will cause a significant performance hit, but apparently some Ubuntu users are happy using proprietary Nvidia drivers, even if it means that when they are done playing World of Goo, quitting the game causes the system to hang and they must hard-reset the system. For those users, it may be that nodelalloc is the right solution for now — personally, I would consider that kind of system instability to be completely unacceptable, but I guess gamers have very different priorities than I do.

This is perhaps a better article at explaining:
http://thunk.org/tytso/blog/2009/03/...-file-problem/

H_TeXMeX_H · 03-25-2009, 07:46 AM

Here are Linus Torvalds' comments on this issue:

http://article.gmane.org/gmane.linux.kernel/811654
http://article.gmane.org/gmane.linux.kernel/811700
http://article.gmane.org/gmane.linux.kernel/811699

And there are more important comments on there if you care to check.

mostlyharmless · 03-25-2009, 02:40 PM

Hmm, Mr Torvalds doesn't mince words. Well, my sentiments exactly.

H_TeXMeX_H · 03-25-2009, 03:14 PM

Indeed he seems to always speak his mind in such cases, I enjoy reading his comments

H_TeXMeX_H · 03-27-2009, 02:58 PM

Here's a more complete explanation about the "squabble":
http://www.linuxquestions.org/questi...d-ext4-714904/

mostlyharmless · 03-27-2009, 03:07 PM

Yes, I saw it. Makes me feel like a kid whose parents are fighting.

DragonSlayer48DX · 03-27-2009, 04:16 PM

Quote:

Originally Posted by syg00

Personally I can't understand why people rush into the "latest and greatest" filesystem - it's what holds all your data.
Bragging rights about being on the "bleeding edge" lose their gloss when your data is trashed.

My $AU0.02

Exactly.

What I can't understand is that back in the good old days (anyone remember the Apple II, Commodore 64, or IBM PC jr?), the system wrote the data when the user gave the command, and nobody was concerned over waiting a few seconds. Now today, with CPUs running millions times faster and drives accessing data hundreds of thousands times faster, and people are willing to trade data integrity for performance!

Is it just me, or has the market push for "bigger and faster" over the past quarter-century warped everyone's minds?