xfs
[Top] [All Lists]

Re: Questions about recovery

To: Steve Lord <lord@xxxxxxx>
Subject: Re: Questions about recovery
From: Lauri Ojantakanen <lauri@xxxxxxxxxxxxx>
Date: Fri, 18 May 2001 14:08:27 +0300
Cc: linux-xfs@xxxxxxxxxxx
In-reply-to: <200105171447.f4HElu921491@xxxxxxxxxxxxxxxxxxxx>
Organization: Solid
References: <200105171447.f4HElu921491@xxxxxxxxxxxxxxxxxxxx>
Sender: owner-linux-xfs@xxxxxxxxxxx
My intention is neither to unplug the machine (it would not hurt as I am 
using xfs in my laptop:) nor to leave out proper shutdown procedures. 
However, my situation is such that I have problems with pppd (my machine 
hangs with it ). I just would like to know what to do to minimize the amount 
of damage if I want to study more this pppd problem.

Rgrds,

//lauri

On Thursday 17 May 2001 17:47, Steve Lord wrote:
> > Lauri Ojantakanen wrote:
> > > Hi,
> > >
> > > I am very happy using xfs (converted everything to xfs:). Usually
> > > everythin
> >
> > g
> >
> > > goes fine with the system but there is one kinda scary thing: sometimes
> > > som
> >
> > e
> >
> > > files seem to be getting filled with null's after recovery. I mean that
> > > aft
> >
> > er
> >
> > > one recovery one opera's bookmark file was still there but the contents
> > > of the file was just  '\0'.
> >
> > This has been discussed a few times on the list.  (See the "might have
> > found a bug..." thread from this week).  The short answer is that it's a
> > feature, and specifying synchronous writes at mount time will minimize
> > it, but will also kill your performance.  You can also tune the bdflush
> > daemon to flush writes more often - also reducing performance.  It's a
> > tradeoff.
> >
> > -Eric
>
> I should also add here that the difference between what happens with XFS in
> this scenario and what would happen with ext2 is that with ext2 you
> probably would not see the new file at all - or the old one, but it depends
> on what got synced and what did not. The reason files show up with no data
> in xfs is that the delayed allocation code which reserves space during a
> write call has to bump the inode size - this size is making it out to disk,
> however, the allocation of real extents and the flush of the data is not.
> If delayed allocation was not being used you would get a file with garbage
> read off the disk, since the disk space would be allocated during the write
> call, but the data would still not be written out into it.
>
> Having a journalled filesystem does not in general mean you go pulling
> the plug without shutting the machine down, recovery is there for dealing
> with accidents and crashes, not so you can use the power button to
> shutdown ;-)
>
> Steve

<Prev in Thread] Current Thread [Next in Thread>