Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 13 Feb 2007 09:49:42 -0800 (PST)
From:      John Kozubik <john@kozubik.com>
To:        freebsd-fs@freebsd.org
Subject:   Re: comments on newfs raw disk ?  Safe ? (7 terabyte array)
Message-ID:  <20070213093220.C95571@kozubik.com>
In-Reply-To: <200702130927.l1D9RBn9034761@lurza.secnetix.de>
References:  <200702130927.l1D9RBn9034761@lurza.secnetix.de>

next in thread | previous in thread | raw e-mail | index | archive | help


Friends,


On Tue, 13 Feb 2007, Oliver Fromme wrote:

> Randy Bush wrote:
>  > this has been a wonderfully well-timed thread as i am about
>  > to hack a 4tb array tomorrow afternoon.  the normal spindle
>  > is separate and partitioned to death and newfsed using the
>  > defaults.  with 2gb of ram, i figure 6gb swap just in case
>  > two userland hogs are running at once, e.g. some hog while
>  > background fsck is running.
>
> A bit careful here ...  Background fsck had some issues,
> especially when the machine crashed or is otherwise reset
> while the background fsck is still running.  It resulted
> in corruption that could not be repaired by fsck anymore.
> I don't know if all of those issues have been resolved in
> RELENG_6, but personally I always disable background fsck
> on all of my machines, just to be safe.


Also remember that filling a filesystem to capacity _while_ it is being
snapshotted will lock your system up[1].  I suppose some interesting crash
loops could arise from this bug on a near full filesystem that someone is
unlucky enough to background fsck.

I think that FreeBSD needs to address the default implementation of
background fsck in general.  UFS2 snapshots are dangerous and unstable,
and have been since their introduction in 5.x [2].

Oliver and I and everyone else here knows the dangers of UFS2 snapshots
and background fsck, and it's very telling that Oliver (like myself)
refuses to use them.  I won't touch either of them, despite overwhelming
financial incentive to implement them [3].

But how many innocent sysadmins and less well informed unix engineers in
the world are loading up FreeBSD because of a perceived history of safety
and stability and putting very important data and services on systems,
which _by default_ have a dangerous ticking time bomb on them ?  Are these
people supposed to fall out of the womb knowing that UFS2 snapshots are
unstable and dangerous, and that _4 years later_ they still aren't safe ?

Until well-informed members of this list feel safe and secure with
snapshots and background fsck in general use, I think background fsck
should be disabled by default.


John Kozubik - john@kozubik.com - http://www.kozubik.com


[1] http://lists.freebsd.org/pipermail/freebsd-bugs/2006-January/016703.html

[2] http://lists.freebsd.org/pipermail/freebsd-bugs/2004-July/007574.html

[3] http://www.rsync.net



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20070213093220.C95571>