Ext3 users - Aug 2001 - Re: EXT3 Worries / Its the army brats that worry me!?!

Hi,

On Wed, Aug 01, 2001 at 11:41:46PM -0700, IT3 Stuart B. Tener, USNR-R
wrote:
> 
>             Well I do appreciate the fact that all upgrades and software
> implementations have risk, but, there comes a point when code must be
frozen
> and be anointed as at some level of quality.
Right.  I consider ext3 on 2.4 code-frozen today pending one checkin
of major auditing of error-handling paths, but that doesn't change any
of the normal core code paths.
> possible failures. Just was curious, if you though I would be loading
> software which works yet?
It works for me, it works under severe stress testing for the authors
and internally in Red Hat, and it works under the stress testing that
VA Linux did before installing it on their turnkey storage servers.
> When do you think it will at a point when your
> legal department and marketing department will feel comfortable telling
> people to implement it or at least supplying it to them in a future
release.
Well, it just went out a few days ago as part of the latest Red Hat
beta release, Roswell.  I can't give release dates, but the public
beta follows on from a lot of other internal and external testing so
I've got reasonable faith in the robustness of the ext3 that was
shipped there.
> queasy, but can you say that you estimate it to be x months away from being
> offered up as part of the default Linus kernel?
It is part of 2.4-ac as of yesterday.  I don't expect Linus to take it
before 2.5.
> However, is it as safe as ext2 is given all things being
> equal?
ext2 on 2.2?  No, that code has had far too much exposure and has an
extremely high level of trust as a result.  2.4?  Less of a difference
--- 2.4's VM and ext2 code are both much younger.  ext3 is still more
complex, and newer, and as a result _has_ to be considered more of a
risk purely from a theoretical point of view.  But the testing ext3
has had on 2.4 has gone extremely well, and with its inclusion into
-ac I hope that any lingering bugs will be flushed out quickly (there
are _always_ bugs you don't find in testing.)

Cheers,
 Stephen

On Wed, Aug 01, 2001 at 11:41:46PM -0700, IT3 Stuart B. Tener, USNR-R
wrote:
>             Well I do appreciate the fact that all upgrades and software
> implementations have risk, but, there comes a point when code must be
frozen
> and be anointed as at some level of quality. I am not looking for you to
> send me a 10 page legal document offering a warranty and guarantee of all
> possible failures. Just was curious, if you though I would be loading
> software which works yet? Apparently, you seem to think so, as I saw no
> massive warnings in your response.
Well, here's my take on it.  It is my firm belief that Stephen Tweedie
and the rest of the ext3 filesystem developers (myself included), are
much more conservative about filesystem stability than the developers
of some other filesystem developers.  Specifically, I believe that
ext3 will prove to be far more stable than reseirfs was at the time
when SuSE started shipping it to all of its guinea
pigs^H^H^H^H^H^H^H^H^H^Husers as part of their distribution.  Time
will tell whether or not I'm correct, but the truth is that the ext3
developers tend to be very careful about what we're willing to call
"production ready".

The I/S folks at my company were willing to risk the home directories
of all of our employees long before I was willing to certify it as
"production ready" (they didn't ask me first) --- on the other
hand,
they didn't lose any data, either, and it worked just fine for them,
so one could argue they made the right choice.

That being said though, I would be nervous if ext3 and e2fsprogs were
used in life-critical or mission-critical applications, just simply
because I personally have very high standards for what that means.
Other developers might say, "Sure!  My code is ready for that!".  But
that may very well be an indication of those developer's
over-confidence (or my being overly paranoid), as opposed to any
indication to the relative code quality of ext3 versus other
filesystems.

I can make statements such as, "e2fsck is the only filesystem
consistency checker and repair utility which I'm aware of that has a
full and comprehensive regression test suite", and "I've never
lost
any data with ext3, and I currently trust my laptop and all of my
e-mail to Linux-2.4.7 with ext3 0.9.5."   

You'll have to decide whether or not that's safe enough for your
application.

							- Ted