Re: [Jfs-discussion] replay fs image to reproduce errors? (was: raid5 failed (power outage))

[Dave K. <sh...@au...>]

On Thu, 2006-08-10 at 16:14 +0100, Christian wrote:
> Hi,
> 
> On Mon, 7 Aug 2006, Dave Kleikamp wrote:
> > Can't read the root directory inode.  There still seems to be some major
> > damage to this file system.  I'm not sure that anything can be done to
> > get fsck to fix it from the current state.
> 
> Given that in most cases one cannot reproduce the fs-breakage of the 
> poor fellow who just lost his/her filesystem, would it be possible to 
> implement sth. like the Reiserfs folks are having?
> 
> They're often advised to "debugreiserfs -p /dev/xxx | gzip -c > xxx.gz"
> and then this b0rken fs can be rebuild again with
> "gzip -dc xxx.gz | debugreiserfs -u /dev/image" (debugreiserfs(8)).
> then one could jfs_fsck on the /dev/image and hopefully find out what's
> the real issue here. Only the xxx.gz has to be sent over the net for 
> analysis, not the whole fs itsself.

This is pretty interesting.  I wasn't aware that reiserfs had this tool.
It would be possible, but it would take a volunteer to write it for jfs.

> But I am no fs-expert, just thinking out loud. Maybe it's not even 
> possible due to the differnt fs-structure....
> 
> greetings,
> Christian.
-- 
David Kleikamp
IBM Linux Technology Center